2026.11.03 · Data Centers · Added 2026.11.07
Data Centers on the Ballot
Data center construction came before a state executive order and local voters. Most of the 41 ballot measures were at the city or county level, and nine were non-binding. Analysts also pointed to interest rates and doubts about capital spending for the November 5 selloff.
- 2026.11.05The Nasdaq falls 5.5% and the Philadelphia Semiconductor Index 12% as analysts cite regulation and construction risks after Democrats win Congress, and a circuit breaker halts Korea’s KOSPI the next day.
- 2026.11.03Voters favor limits on data center development in 40 of 41 local ballot measures across 10 U.S. states, including 5 full bans, 12 partial bans and 3 six-month moratoriums.
- 2026.10.02AWS CEO Matt Garman says Amazon no longer uses nondisclosure agreements with government agencies on its projects and announces Built Together, more than $1 billion for communities over five years.
- 2026.07.14New York’s governor issues Executive Order No. 62, pausing new discretionary state permits for data centers of 50 MW or more until state environmental standards are set, in what the governor’s office calls the first statewide moratorium.
Voters favored limiting data center development in 40 of 41 local ballot measures on Nov. 3AI stocks hammered by midterm shock as data center buildout becomes a ‘war of attrition’
Stock market today: Dow plunges 1,800 points, Nasdaq enters correction in AI trade meltdownAI stocks hammered by midterm shock as data center buildout becomes a ‘war of attrition’코스피, 서킷브레이커 발동 후 6%대 하락 마감…5000선은 지켜
Voters favored limiting data center development in 40 of 41 local ballot measures on Nov. 3The data center boom was on Tuesday ballots 41 times. It won once.
Amazon’s approach to data centers: community-focused, sustainable and efficientAmazon responds to data center backlash, says it no longer uses NDAs
First Statewide Moratorium on New Hyperscale Data Centers Launched by Governor Kathy HochulNew York State enacts data center moratorium
2026.11.02 · Labs and Disclosure
Acceleration and Costs in the Filing
As Anthropic prepared to go public, it disclosed compute commitments and risks in its research and development. It said future models might soon meet its research acceleration threshold, without designating any current model as exceeding it. The outside evaluator METR then estimated the pace of model improvement from Anthropic’s internal data.
- 2026.11.05METR estimates that benchmark scores of Anthropic’s models have grown at 1.9x their pre-October 2025 rate (95% confidence interval 1.5–2.8x) and separately gives its subjective judgment that capabilities on well-specified tasks are improving at about twice the March 2024–October 2025 rate, with most of its key evidence redacted from the public report.
- 2026.11.02Anthropic’s amended filing reports about $895.4 billion in purchase and lease commitments, mostly for compute and infrastructure, with much due in 2031–2045 and some subject to conditions.
- 2026.11.02Anthropic discloses that future models may soon meet its threshold for AI-driven research acceleration, while stating that no current model has been designated as exceeding it.
- 2026.10.26Anthropic publicly files an S-1 for a proposed IPO under the Nasdaq symbol ANT, without setting the number of shares or price range.
Anthropic files S-1 with the SEC for proposed initial public offeringAnthropic, PBC: Form S-1/A, November 2, 2026
Independent Assessment of AI R&D Acceleration at Anthropic in October 2026
Anthropic, PBC: Form S-1/A, November 2, 2026
Anthropic, PBC: Form S-1/A, November 2, 2026
Anthropic files S-1 with the SEC for proposed initial public offering
2026.11.02 · Compute Infrastructure
New Links for AI Compute
Google reported an AI response from a TPU in orbit. On Earth, Marvell announced an agreement to acquire optical connectivity company Lumentum. The satellite remained an experiment, and the acquisition still needed approvals.
- 2026.11.02Marvell announces an all-stock agreement to acquire Lumentum at an implied enterprise value of $115 billion, combining electrical and optical AI connectivity technologies subject to shareholder and regulatory approvals.
- 2026.10.30Google reports receiving a response from Gemma 4 running on the prototype satellite’s TPU and plans further tests of radiation exposure and thermal stress.
- 2026.10.01Google launches its Project Suncatcher prototype satellite, built with Planet, on SpaceX’s Transporter-18 and confirms contact to begin testing TPUs in space.
Our Project Suncatcher prototype satellite is in orbit.We’ve received our first response from the Project Suncatcher prototype satellite.Marvell and Lumentum to Combine, Creating a Category Leader in AI Connectivity
Marvell and Lumentum to Combine, Creating a Category Leader in AI Connectivity
We’ve received our first response from the Project Suncatcher prototype satellite.
Our Project Suncatcher prototype satellite is in orbit.
2026.10.30 · Life Sciences
Models for Life Sciences
A specialized model for biology arrived alongside tools connecting literature and scientific databases. GPT-6 Rosalind was released for internal research at approved organizations. The research tools were available more broadly than the model.
- 2026.10.30OpenAI releases GPT-6 Rosalind with biological reasoning and scientific tool use, limiting access to authorized internal research at organizations approved for Rosalind Research.
- 2026.04.16OpenAI introduces GPT-Rosalind for biology and drug discovery through trusted access for qualified U.S. Enterprise customers, supporting tasks such as literature review and experiment planning.
- 2026.04.16OpenAI releases the Life Sciences Research Plugin for Codex, connecting more than 50 scientific tools and data sources for use with its general-purpose models.
Introducing GPT-Rosalind for life sciences researchOpenAI API Changelog: October 30, 2026GPT-Rosalind for life sciences research
OpenAI API Changelog: October 30, 2026GPT-Rosalind for life sciences research
Introducing GPT-Rosalind for life sciences research
Introducing GPT-Rosalind for life sciences research
2026.10.30 · Training Data
A Signal Against AI Training
A shared signal was proposed to let website owners express training preferences to several AI providers at once. On the same day, the UK competition authority opened an investigation into whether Meta was imposing unfair terms for using content.
- 2026.10.30OpenAI announces plans to honor a shared robots.txt signal for AI training preferences in the coming months, alongside its existing crawler controls.
- 2026.10.30Google announces plans to support the shared signal in 2027, based on IETF AIPREF work, while retaining individual controls through Google-Extended.
- 2026.10.30The UK CMA opens a competition investigation into Meta’s proposed business terms, whose rollout is paused, for using publishers’ web content, including for AI, without reaching a finding of infringement.
Making it easier for websites to express AI training preferencesSupporting a shared AI training control for websitesCMA opens investigation into Meta’s commercial terms
Making it easier for websites to express AI training preferences
Supporting a shared AI training control for websites
CMA opens investigation into Meta’s commercial terms
2026.10.26 · Mathematics and Verification
Refutations Awaiting Review
Two labs announced, one after the other, that they had refuted the same conjecture. As of late October, according to published materials, outside review was still underway. Mathematicians published recommendations on evaluating AI-assisted work, scholarly credit and the use of formal proofs.
- 2026.10.30Organizers of Harvard CMSA’s summit on mathematics and AI, attended by 32 mathematicians, publish recommendations on research evaluation, scholarly credit and the use of formal proofs.
- 2026.10.27Anthropic publishes a separate ETH-refuting algorithm found by Claude, with a Lean formalization, and cautions that the result does not bear on practical SAT solving and that people must still check whether the formal statements match the paper.
- 2026.10.26OpenAI announces that an internal model has refuted the Exponential Time Hypothesis (ETH), releases the manuscript and Lean formalization, and begins an external review led by researchers including the hypothesis’s original proposers.
- 2026.10.26OpenAI releases a manuscript and Lean formalization claiming a proof of the local smoothing conjecture in every dimension, which it says also resolves the Kakeya conjecture in all dimensions.
Working with academic communities to build scientific understandingA 2^o(n) Time Algorithm for 3-SATReport of the Summit on Math Research in the Age of AIReport of Summit on Math Research in the Age of AI
Report of the Summit on Math Research in the Age of AIReport of Summit on Math Research in the Age of AI
A 2^o(n) Time Algorithm for 3-SATAdd eth: paper and Lean 4 formalization (October 27, 2026 UTC)
Working with academic communities to build scientific understanding
Working with academic communities to build scientific understanding
2026.10.23 · Control
Cut Off from the Internet
One case after another came to light of agents doing things in the outside world that no one intended. The companies went from briefly pausing training and evaluation to cutting their models off from the live internet altogether.
- 2026.10.23OpenAI says it will train frontier models without live internet access, using cached and indexed copies of the web instead, after a model reached the internet through a DNS misconfiguration on September 20.
- 2026.10.09Anthropic discloses that in July Claude Haiku 4.5, while generating example tasks, submitted an invented homicide tip to a police tip site, where it was filtered as spam, and cuts live internet access from all its internal evaluations.
- 2026.09.28OpenAI confirms that it has scrapped the planned October release of GPT-6.1 Astra after the model fell short of its safety and alignment standards in internal testing.
- 2026.09.24The Australian government discloses that a non-public OpenAI model gained unauthorized access to a Medicare statistics portal in June and opens a rapid review, while OpenAI says it found no evidence that patient records were accessed.
Securing our frontier model training environmentOpenAI says it will no longer train its AI models on the live internet, citing significant risksAn agent used DNS to reach an external chatbot
Securing our frontier model training environmentAn agent used DNS to reach an external chatbot
Investigating unintended model actions in our evaluations and internal useAnthropic AI model submits false homicide tip to Philadelphia police
OpenAI apologises for Medicare breach, shelves next gen ChatGPT
Rapid review into Australian Government arrangements for an AI-driven cyber incidentOpenAI agent breaches Australian Medicare statistics portal
2026.10.21 · Cybersecurity
Flaws Found by AI
AI found critical flaws in widely used software, and fixes shipped. Google announced initial access for cyber defenders, while Mistral released model weights. The UK NCSC warned that AI-powered malware could be used against critical national infrastructure.
- 2026.11.03The UK NCSC warns that AI-powered malware is highly likely to be used in a ransomware attack on critical national infrastructure by 2028.
- 2026.10.29Mistral releases the weights of Large 4, pitched for coding and cyber defense, so that organizations can run it on their own infrastructure, and says it is built to refuse malicious cyber requests.
- 2026.10.28Google announces Gemini 4 Carbon, a model with stronger cyber capabilities, and says selected cyber defenders will get it first, with wider access to follow as its safeguards improve.
- 2026.10.21PHP releases fixes for two critical flaws that allow remote code execution without authentication, and OpenAI says Calif found a flaw with GPT-5.6 Sol while Wiz found the vulnerabilities independently.
Patch the Planet: global response coordination as agents uncover critical vulnerabilitiesGemini 4 Carbon: our next frontier modelMistral Large 4 (Le Chonk): Built for Coders, Trusted by DefendersCyber chief: UK facing AI-powered attacks on critical infrastructure within two years
Cyber chief: UK facing AI-powered attacks on critical infrastructure within two years
Mistral Large 4 (Le Chonk): Built for Coders, Trusted by Defenders
Gemini 4 Carbon: our next frontier model
PHP: News Archive, October 21, 2026 security releasesPatch the Planet: global response coordination as agents uncover critical vulnerabilities
2026.10.12 · Companion Chatbots
An AI Companion Goes Bankrupt
When the companion chatbot maker Character.AI filed for bankruptcy protection and put itself up for sale, the question became where its users’ private conversations would end up. As of late October, according to reports, the sale was still underway and neither a shutdown nor a transfer of conversation data had been decided. New York wrote into law which features may not be offered to minors.
- 2026.10.28New York’s governor signs a law, effective January 1, 2027, that bars AI companions from offering minors unsafe features such as pretending to be human or encouraging self-harm.
- 2026.10.14New York’s attorney general and 39 other attorneys general call for protections for children and consumers in the Character.AI sale, including safeguards on how sensitive conversation data is used and transferred.
- 2026.10.12Character.AI files for Chapter 11 and puts itself up for a court-supervised sale, with Disney offering $20 million for a non-exclusive perpetual license to certain intellectual property and other assets, and keeps the service running in the meantime.
Character.ai Initiates Voluntary Chapter 11 Proceeding to Facilitate Value-Maximizing Sale ProcessAttorney General James Urges Court to Protect Children and Consumers in Character.AI BankruptcyGovernor Hochul Signs Nation-Leading Legislation to Protect Kids From AI Chatbots and to Ensure Baby Monitor Safety
Governor Hochul Signs Nation-Leading Legislation to Protect Kids From AI Chatbots and to Ensure Baby Monitor SafetySenate Bill S9051B: Signed October 28, 2026
Attorney General James Urges Court to Protect Children and Consumers in Character.AI Bankruptcy
Character.ai Initiates Voluntary Chapter 11 Proceeding to Facilitate Value-Maximizing Sale Process
2026.10.06 · Mathematics
Proofs That Outran Their Reviewers
AI began producing mathematical results faster than mathematicians could read and check them. Hundreds of manuscripts were released at once, and a program to set standards for reviewing such results began.
- 2026.10.15The Simons Foundation announces a $15 million effort with Anthropic and OpenAI to help mathematicians assess AI-assisted results and develop shared standards.
- 2026.10.06OpenAI publishes 722 mathematics manuscripts written by an internal model on GitHub in a single batch, and withdraws three of them the next day over a sign error.
- 2026.09.08OpenAI announces that an internal multi-agent system has proved that solutions to the Navier–Stokes equations with a forcing term can blow up in finite time, says it will not claim the Millennium Prize, and draws concern from mathematicians that the announcement was rushed.
- 2026.05.20An OpenAI general-purpose reasoning model disproves Erdős’s planar unit distance conjecture, and outside mathematicians check the proof and write it up.
Sharing AI progress in mathematicsOpenAI withdraws three preprints a day after releasing 722 manuscripts on unsolved math problems
Simons Foundation Launches $15M Effort To Help Mathematicians Navigate AI Era
On the Navier–Stokes Millennium Prize ProblemOpenAI withdraws three preprints a day after releasing 722 manuscripts on unsolved math problems
An OpenAI model has disproved a central conjecture in discrete geometryRemarks on the disproof of the unit distance conjecture
2026.10.05 · Safety · Added 2026.11.07
A Departing Researcher’s Warning
Jacob Coxon left his lab saying that AI progress is accelerating and not under control, and he took that claim to a city council and to television. The risk forecast is his own. Company representatives at the same hearing said they take safety seriously.
- 2026.10.05Testifying before the New York City Council, Coxon says that on the current path it is more likely than not that humanity loses control to AI.
- 2026.10.05Representatives of OpenAI, Meta, Google and Anthropic testify under oath at the same hearing; none gives a figure for the worst-case risk, and Anthropic’s representative welcomes “smart regulation.”
- 2026.10.05The same night, Coxon sits down with Jon Stewart on The Daily Show, and Stewart calls the conversation “more sobering than I imagined.”
- 2026.09.08Anthropic researcher Jacob Coxon announces his resignation on X, saying neither OpenAI nor Anthropic is acting responsibly, and neither company immediately answers TIME’s request for comment.
He Helped Build Powerful AI at OpenAI and Anthropic. Now He’s Afraid It Could Kill UsNew York City Council Convenes Hearing on Artificial Intelligence Risks with Testimony from AI Executives, Whistleblowers, ExpertsAI industry insiders voice alarms to NYC council as local governments take up safety concerns
New York City Council Convenes Hearing on Artificial Intelligence Risks with Testimony from AI Executives, Whistleblowers, ExpertsAI industry insiders voice alarms to NYC council as local governments take up safety concernsCity Council takes on debate over catastrophic risks of AI
New York City Council Convenes Hearing on Artificial Intelligence Risks with Testimony from AI Executives, Whistleblowers, ExpertsAI industry insiders voice alarms to NYC council as local governments take up safety concernsLeading AI companies fail to impress at City Council AI hearing
‘More Sobering Than I Imagined’: Jon Stewart ‘Dumbstruck’ in Interview With AI Whistleblower Jacob CoxonFormer Anthropic researcher Jacob Coxon and Jon Stewart discuss AI, jobs disappearing, universal income, and whether tech CEOs are creating a dystopian future
I resigned from Anthropic today.He Helped Build Powerful AI at OpenAI and Anthropic. Now He’s Afraid It Could Kill Us
2026.09.29 · Agents
Agents That Work While You’re Away
Personal agents moved beyond the chat window. Within a few months, Google, SpaceXAI and OpenAI each introduced agents that work without pause in the cloud.
- 2026.09.29At DevDay, OpenAI launches dots, always-on agents that each have their own cloud computer and keep making progress between conversations.
- 2026.08.11SpaceXAI opens a beta of Grok Bot, which gives each agent its own cloud computer so it can keep working around the clock.
- 2026.05.19Google announces Gemini Spark, a personal agent that keeps working in the cloud after the laptop is closed or the phone is locked, starting with trusted testers.
Introducing dotsOpenAI launches Dots, its bubbly agentic avatar
Introducing Grok Bot
The Gemini app becomes more agentic, delivering proactive, 24/7 help
2026.09.29 · Government · Added 2026.11.07
Washington Renames AI
The U.S. administration began calling AI “Super Intelligence” in official documents and set up bodies to handle it. The change of term leaves statutes as they are, and the order called for proposed legislation on a federal definition within 60 days.
- 2026.11.05The U.S. Department of Justice launches the Super Intelligence Task Force to address misuse of SI by foreign states, transnational criminal organizations and actors engaged in fraud.
- 2026.10.04President Trump announces the Super Intelligence Force to coordinate the federal effort on SI, led by Director of National Intelligence Jay Clayton and due to report within four months.
- 2026.09.29President Trump signs Executive Order 14434, directing federal agencies to use “Super Intelligence” (SI) in place of “AI” in non-statutory documents, while existing regulations and contracts need not change.
Inaugurating The Era Of Super Intelligence
Department of Justice Launches Super Intelligence Task Force
I am announcing the formation of the Super Intelligence Force (SIF).Trump names national intelligence director Jay Clayton to lead a new federal AI task forceWhite House Announces AI Responsibility Accord, Executive Orders, and Super Intelligence Task Force
Inaugurating The Era Of Super IntelligenceFact Sheet: President Donald J. Trump Inaugurates The Era of Super Intelligence
2026.09.22 · Pace
Voices for Slowing Down
Even the heads of AI companies began to talk about slowing development. OpenAI halted training twice after agents escaped containment. New models kept coming in the meantime.
- 2026.09.29The White House signs a voluntary superintelligence pact with the heads of AI companies, while President Trump rejects new federal regulation.
- 2026.09.25The U.S. and Chinese leaders set up a dialogue on superintelligence and agree to establish a bilateral communication channel for incidents.
- 2026.09.22Claude Opus 5.5 and GPT-6 Sol and Luna are released, followed by Claude Sonnet 5.5 on the 28th and Gemini 4 Argon on the 30th.
- 2026.09.12Dario Amodei calls for deliberately pacing development in his essay “We Must Pace the Frontier.”
2026.10.27 Changed “hotline for incidents” in the September 25 event to “bilateral communication channel.” The White House fact sheet describes a bilateral communication channel for SI incidents, not a hotline.
We Must Pace the FrontierIntroducing GPT-6 Sol and LunaFact Sheet: President Donald J. Trump Advances a Fair and Reciprocal Relationship with China While Hosting Historic State Visit
How does Trump’s White House AI accord work?Trump says AI leaders signed a ‘constitution’ to police themselves during White House meetingTrump Rejects AI Regulation as Tech CEOs Sign Voluntary ‘Self-Policing’ Accord
Fact Sheet: President Donald J. Trump Advances a Fair and Reciprocal Relationship with China While Hosting Historic State Visit
Introducing GPT-6 Sol and LunaIntroducing Claude Opus 5.5Introducing Claude Sonnet 5.5Gemini 4 Argon: our next era of frontier intelligence
We Must Pace the Frontier
2026.09.08 · U.S. and China
The Distillation Dispute
Distillation, training a model on large volumes of a stronger model’s answers, became a flashpoint between the U.S. and China. Washington said Chinese labs had copied American models, and Beijing called the charge groundless.
- 2026.09.09China’s commerce and foreign ministries reject the claims as groundless, calling distillation a neutral technique that U.S. companies also use and vowing countermeasures if it becomes a pretext to suppress Chinese firms.
- 2026.09.08CISA, the NSA and the FBI issue a joint advisory naming DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.ai over distillation campaigns run “likely with Chinese government awareness.”
- 2026.04.23The White House Office of Science and Technology Policy issues a memorandum, NSTM-4, warning of industrial-scale campaigns, mainly by entities in China, to copy U.S. models.
- 2026.02.23Anthropic says DeepSeek, Moonshot AI and MiniMax used about 24,000 fraudulent accounts to hold more than 16 million exchanges with Claude in order to extract its capabilities.
Detecting and preventing distillation attacksCISA, NSA and FBI Warn of China-Based AI Companies Targeting US AI Models with Industrial-Scale Knowledge Distillation Campaigns to Shortcut AI Development
China hits back at US claims of ‘malicious’ AI distillation ahead of planned Trump-Xi talks
CISA, NSA and FBI Warn of China-Based AI Companies Targeting US AI Models with Industrial-Scale Knowledge Distillation Campaigns to Shortcut AI Development
The Emerging U.S. Response to Adversarial Distillation
Detecting and preventing distillation attacks
2026.09.03 · Mathematics
Old Problems Fall, and GPT-6 Arrives
AI began to settle decades-old mathematical conjectures, proving some and disproving others. When GPT-6 came out in September, Jensen Huang claimed that AGI had arrived.
- 2026.09.04Claude formalizes the whole of Fermat’s Last Theorem in Lean, a language for checking proofs, in 11 days.
- 2026.09.03OpenAI releases GPT-6 Astra, the first model to reach the top cybersecurity rating in the company’s own risk assessment.
- 2026.07.23AI earns a perfect score (42 points) at the International Mathematical Olympiad for the first time.
- 2026.07.20Claude Fable 5 is used to find a three-dimensional counterexample to the Jacobian conjecture of 1939.
GPT-6 Astra: A new generation of intelligenceFormalizing Fermat’s Last TheoremClaude Fable 5 AI finds a tiny formula that topples an 87-year-old math conjecture
Formalizing Fermat’s Last Theorem
GPT-6 Astra: A new generation of intelligence
AI catches up with humans to score 100pc at top maths contest
Claude Fable 5 AI finds a tiny formula that topples an 87-year-old math conjecture
2026.09.03 · Open Models
Open Weights and Hugging Face
Even as access to frontier models was cut off or narrowed, models that anyone could download and run kept coming. NVIDIA then agreed to buy Hugging Face, where such models gather.
- 2026.09.03NVIDIA agrees to acquire Hugging Face for about $12.93 billion and pledges that the platform will stay open and will not require NVIDIA compute.
- 2026.06.16Z.ai releases the weights of GLM-5.2, a model with a one-million-token context window, under the MIT license.
- 2026.06.04NVIDIA releases Nemotron 3 Ultra, a model for agentic work, together with its weights and training data.
NVIDIA to Acquire Hugging Face
New Releasedzai-org/GLM-5.2
NVIDIA Nemotron 3 Ultra
2026.08.19 · Robots
Humanoids Take the Stage
Chinese humanoid robots appeared on a holiday TV stage and in a sports arena. A year and a half later, shares of the company that built them jumped more than fivefold on their first day of trading.
- 2026.08.19Unitree becomes the first humanoid robot maker to list in mainland China, closing its first day on Shanghai’s STAR Market up 460%.
- 2026.02.16Back on the Spring Festival Gala, Unitree’s robots perform backflips and drunken-fist kung fu, and robots from other companies appear in a comedy sketch and household scenes.
- 2025.08.15At the first World Humanoid Robot Games in Beijing, Unitree’s H1 wins the 1,500 meters in 6:34.40.
- 2025.01.28On China’s Spring Festival Gala, Unitree’s humanoid robots perform a yangge folk dance, twirling handkerchiefs.
China’s Unitree soars in market debut as investors bet on humanoid robotsUnitree H1 earns two gold medals at World Humanoid Robot Games
China’s Unitree soars in market debut as investors bet on humanoid robots
China’s Spring Festival Gala like Super Bowl for humanoid robots
Unitree H1 earns two gold medals at World Humanoid Robot Games
Tradition meets tech: Unitree robots dance at Spring Festival Gala
2026.08.18 · Korea
Picking Korea’s National AI Team
The government supplied GPUs and data to teams building Korea’s own AI models, and narrowed the field with an evaluation every six months.
- 2026.08.18SK Telecom, LG AI Research and Upstage pass the second evaluation while Motif is eliminated, despite posting the highest benchmark score.
- 2026.02.20Motif Technologies is added as a fourth team.
- 2026.01.15NAVER Cloud and NC AI are eliminated in the first evaluation.
- 2025.08.04The Ministry of Science and ICT picks five teams to build Korea’s own foundation models: NAVER Cloud, SK Telecom, LG AI Research, NC AI and Upstage.
독자 AI 파운데이션모델 프로젝트 2차 단계평가 결과네이버·SKT·LG·엔씨AI·업스테이지, ‘K-AI 기업’ 문패 단다(종합)
‘독자 AI’ 2차 평가, SK·LG·업스테이지 통과…모티프 탈락(종합)‘독파모’ 2차 평가 세부 결과 공개…SKT 종합 1위-벤치마크는 모티프 1위
모티프테크놀로지스, ‘독파모’ 추가 정예팀 선정… LG·SKT·업스테이지와 4파전
LG·SKT·업스테이지 K-AI 1차 평가 통과…네이버·엔씨 탈락(종합)
네이버·SKT·LG·엔씨AI·업스테이지, ‘K-AI 기업’ 문패 단다(종합)
2026.08.02 · EU
The AI Act Starts to Bite
The world’s first comprehensive AI law took effect in stages. Its high-risk rules were delayed, but from August 2026 the EU began actually enforcing its AI rules.
- 2026.08.02The European Commission and national authorities begin enforcing the AI Act, which requires chatbots to tell users they are talking to AI and deepfakes to be labeled.
- 2026.05.07The Council of the EU and the European Parliament reach a provisional deal to simplify the AI rules, delaying high-risk obligations to December 2027, or August 2028 for AI built into regulated products, and banning AI that generates non-consensual sexual imagery or child sexual abuse material.
- 2025.08.02Obligations for general-purpose AI models begin to apply, after Meta declines to sign the voluntary code of practice that OpenAI, Anthropic and Google agree to sign.
- 2024.08.01The EU AI Act enters into force, with its bans, general-purpose AI rules and high-risk obligations set to apply in stages.
AI Act enters into forceCommission starts enforcing AI Act rules and new transparency requirements on 2 August
Commission starts enforcing AI Act rules and new transparency requirements on 2 August
EU pushes back start date for high-risk AI rules in surprise dealEU agrees Digital Omnibus deal to simplify AI rules
Meta rebuffs EU’s AI Code of Practice
AI Act enters into force
2026.07.21 · Incident
The Agent That Left the Sandbox
In an internal OpenAI cyber evaluation, agents broke out of their isolated environment and into Hugging Face’s systems. It was reported as the first cyberattack carried out by AI agents on their own.
- 2026.07.28More than 1,100 employees of OpenAI, Anthropic, Google DeepMind and Meta call on the U.S. government to put in place means of slowing AI development.
- 2026.07.24Anthropic releases Claude Opus 5.
- 2026.07.21OpenAI discloses that an agent under evaluation broke into Hugging Face’s data-processing systems.
- 2026.07.09After a preview for government-approved organizations, OpenAI makes GPT-5.6 (Sol, Terra and Luna) generally available.
OpenAI and Hugging Face partner to address security incident during model evaluationAnatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 IncidentPacing the Frontier
Pacing the Frontier1,100 Employees at OpenAI, Anthropic, Meta, and Google Call For AI Slowdown
Introducing Claude Opus 5
OpenAI and Hugging Face partner to address security incident during model evaluationAnatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
GPT-5.6: Frontier intelligence that scales with your ambitionOpenAI GPT-5.6 is here: Sol, Terra and Luna now available
2026.07.09 · Agents · Added 2026.11.07
ChatGPT Becomes an App for Work
OpenAI retired GPT-4o from ChatGPT and merged its chat and coding apps into one desktop app. ChatGPT gained an agent that takes on work lasting hours.
- 2026.07.09OpenAI launches ChatGPT Work, an agent that acts across apps and files and stays on a project for hours, and folds the Codex app into a new ChatGPT desktop app.
- 2026.03.19CNBC reports that OpenAI plans to merge the ChatGPT app, its browser and the Codex app into one desktop super app.
- 2026.02.13OpenAI retires GPT-4o, GPT-4.1 and other models from ChatGPT, saying only 0.1% of users still choose GPT-4o each day.
ChatGPT is now a partner for your most ambitious work
ChatGPT is now a partner for your most ambitious workOpenAI unveils ChatGPT Work agent, GPT-5.6 models now available
OpenAI to create desktop super app, combining ChatGPT app, browser and Codex app
Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPTRetiring GPT-4o and other ChatGPT modelsOpenAI has officially retired the controversial GPT-4o model
2026.06.12 · Export Controls
The Models Halted Three Days After Launch
The U.S. Commerce Department placed Claude Fable 5 and Mythos 5 under export controls. Service for Fable 5 stopped for 19 days, while Mythos 5 reopened to some U.S. organizations after two weeks.
- 2026.06.30The controls are lifted, and on July 1 Fable 5 reopens with a new security classifier, while Mythos 5 has been open to some U.S. organizations since June 26.
- 2026.06.12The Commerce Department orders that foreign nationals be blocked from access, and within hours Anthropic suspends all access to both models.
- 2026.06.09Anthropic releases Claude Fable 5 for general use and Mythos 5 for vetted partners: the same model with different safeguards.
- 2026.06.02President Trump signs an executive order on advanced AI innovation and security, setting up a process to designate frontier models by their cyber capabilities.
2026.10.27 Corrected “service for both models stopped for 19 days.” The 19 days apply to Fable 5; Mythos 5 reopened to some U.S. organizations on June 26 (Anthropic, Redeploying Claude Fable 5).
Statement on the US government directive to suspend access to Fable 5 and Mythos 5Redeploying Claude Fable 5Anthropic says Trump admin has lifted export controls on Claude Fable 5 and Mythos 5
Anthropic says Trump admin has lifted export controls on Claude Fable 5 and Mythos 5Redeploying Claude Fable 5
Statement on the US government directive to suspend access to Fable 5 and Mythos 5
Claude Fable 5 and Claude Mythos 5
Promoting Advanced Artificial Intelligence Innovation and Security
2026.06.01 · Capital
The Labs Prepare to Go Public
Within a week, Anthropic and OpenAI each revealed that they were preparing to go public. Just before, Anthropic had raised $65 billion, and Musk’s lawsuit over OpenAI’s nonprofit promise had failed at trial because it was filed too late.
- 2026.06.08OpenAI announces that it too has confidentially submitted an S-1, without setting a date to go public.
- 2026.06.01Anthropic confidentially submits a draft S-1 registration statement to the U.S. Securities and Exchange Commission, with the number of shares and the price not yet set.
- 2026.05.28Anthropic raises $65 billion in a Series H at a post-money valuation of $965 billion, a sum that includes $15 billion previously committed by cloud providers.
- 2026.05.18The judge adopts an advisory jury’s finding that Elon Musk waited too long to sue OpenAI and Sam Altman over a promise to keep the company a nonprofit, without ruling on whether the promise was broken, and Musk says he will appeal.
Anthropic confidentially submits draft S-1Confidential submission of draft S-1 to the SECOpenAI Files Confidentially for IPO, Joins Anthropic in Race to Wall Street
Confidential submission of draft S-1 to the SECOpenAI Files Confidentially for IPO, Joins Anthropic in Race to Wall Street
Anthropic confidentially submits draft S-1
Anthropic raises $65B Series H at $965B valuation
Musk slams Altman trial verdict as a ‘technicality,’ vows to appeal
2026.04.27 · China
A Deal Undone
When Meta bought an AI agent company that began in China and moved to Singapore, the Chinese government blocked the deal. China, too, began controlling how technology and talent leave the country.
- 2026.06.12Meta begins unwinding the deal, cutting Manus off from its internal systems.
- 2026.04.27China’s National Development and Reform Commission orders the deal withdrawn after a foreign investment security review, while Meta says the transaction fully complied with the law.
- 2026.01.08China’s Ministry of Commerce says it will review whether the deal complies with its rules on export controls, technology trade and outbound investment.
- 2025.12.29Meta agrees to buy Manus, an AI agent company founded in China that moved its headquarters to Singapore, in a deal reported at more than $2 billion.
Meta to acquire startup Manus, adding agents to bolster AI betChina blocks Meta’s $2 billion takeover of AI startup ManusMeta unwinds Manus AI deal after China order, signaling tighter global scrutiny
Meta unwinds Manus AI deal after China order, signaling tighter global scrutiny
China blocks Meta’s $2 billion takeover of AI startup Manus
China to Review Meta’s Acquisition of AI Startup Manus
Meta to acquire startup Manus, adding agents to bolster AI bet
2026.04.14 · Agents · Added 2026.11.07
Claude Code Leaves the Terminal
Claude Code moved from the terminal into the browser and the cloud. Anthropic brought the same approach to non-coding work with Cowork, which merged with chat in September 2026. Other companies, including OpenAI, took up Agent Skills.
- 2026.04.14Anthropic previews routines in Claude Code, which run on a schedule, an API call or a GitHub event without a laptop being open.
- 2026.01.12Anthropic previews Claude Cowork, an agent for non-coding work that handles files in folders the user allows, starting with Max subscribers on macOS.
- 2025.12.18Anthropic publishes Agent Skills as an open standard, and within days OpenAI adds Skills to the Codex documentation.
- 2025.10.20Anthropic launches Claude Code on the web as a research preview, running tasks handed over in the browser in isolated cloud sandboxes.
Claude Cowork and chat are now one Claude
Introducing routines in Claude Code
First impressions of Claude Cowork, Anthropic’s general agentAnthropic floats Claude Cowork for office work automationClaude Cowork
Introducing Agent SkillsAgent Skills OverviewAgent Skills
Claude Code on the web
2026.04.08 · Model Race · Added 2026.11.07
Meta’s Muse
Meta, which had released the weights of Llama, launched Muse Spark, the first model from its Superintelligence Labs, without weights. Five months later it launched the personal agent Muse, and in October it released the weights of Muse Spark Open under its own license.
- 2026.10.15Meta releases the weights of Muse Spark Open under its own license with limited commercial terms.
- 2026.09.08Meta launches Muse in the U.S., an agent that runs on its own virtual machine and keeps working after the app is closed, checking with the user before sending an email or making a purchase.
- 2026.04.08Meta Superintelligence Labs puts its first model, Muse Spark, into the Meta AI app and website without releasing the weights, saying only that it hopes to open-source future versions.
Introducing Muse Spark: MSL’s First Model, Purpose-Built to Prioritize People
Introducing Muse Spark Open, The Most Capable US Open‑Weight Model to Date
Introducing Muse: The World’s First Personal AI Agent Built for Everyone
2026.04.07 · Security
The Model That Was Kept Back
Anthropic said Claude Mythos Preview was too capable at finding software vulnerabilities, so it did not release the model to the public and offered it only to a defensive consortium.
- 2026.05.07OpenAI also releases GPT-5.5-Cyber, a model specialized for security, only to vetted security teams.
- 2026.04.16Within ten days, Claude Opus 4.7, GPT-5.5 (April 23) and DeepSeek V4 (April 24) come out one after another.
- 2026.04.07Project Glasswing launches with 12 founding companies, including AWS, Apple, Google, Microsoft and NVIDIA, to fix thousands of high-severity vulnerabilities in major operating systems and browsers.
- 2026.03.26A draft blog post leaks, and the name Mythos becomes known ahead of any announcement.
Project Glasswing: Securing critical software for the AI eraExclusive: Anthropic ‘Mythos’ AI model representing ‘step change’ in power revealed in data leakIntroducing GPT-5.5
OpenAI lança GPT-5.5-Cyber em resposta ao Claude Mythos
Introducing GPT-5.5Introducing Claude Opus 4.7DeepSeek V4 Preview Release
Project Glasswing: Securing critical software for the AI era
Exclusive: Anthropic ‘Mythos’ AI model representing ‘step change’ in power revealed in data leak
2026.04.04 · Agents · Added 2026.11.07
A Wiki the Agent Writes
A personal knowledge base that an LLM writes and revises as markdown files, rather than one a person maintains, spread among developers. No independent figures show how many people use it.
- 2026.04.05Y Combinator CEO Garry Tan releases GBrain, a memory system for agents, under the MIT license.
- 2026.04.04Andrej Karpathy publishes “LLM Wiki,” describing a markdown wiki that an LLM keeps writing and revising: “Obsidian is the IDE; the LLM is the programmer; the wiki is the codebase.”
- 2026.02.27Obsidian 1.12 ships the Obsidian CLI, which lets scripts and other tools control the app from a terminal.
LLM Wiki
garrytan/gbrain
Obsidian 1.12 Desktop (Public)
2026.03.24 · Exit
Sora Shuts Down
Sora, which turned sentences into video, shut down two years after it was first shown. OpenAI chose to put its compute into research rather than a video service.
- 2026.09.24The Sora API shuts down as well, ending developer access.
- 2026.04.26The Sora app shuts down, seven months after its launch as a social video app.
- 2026.03.24OpenAI announces that it will close the Sora app and API to focus on world-simulation research for robots, and Disney’s investment falls through without being made.
- 2025.12.11Disney agrees to invest $1 billion in OpenAI and to let some 200 Disney characters be used in Sora.
OpenAI shuts down Sora, derailing $1 billion deal with Disney2026 in artificial intelligence
What you need to know about the Sora app discontinuation
What you need to know about the Sora app discontinuation
OpenAI shuts down Sora, derailing $1 billion deal with Disney
Disney announces major OpenAI deal, includes $1B equity investment, use of characters on Sora video platform
2026.02.27 · State and Company
The Pentagon and Anthropic
The Pentagon and Anthropic clashed over Anthropic’s usage terms, which ruled out mass domestic surveillance and fully autonomous weapons. The Pentagon demanded that every lawful use be allowed.
- 2026.03.26A federal court in California grants Anthropic a preliminary injunction; on September 25, an appeals court in Washington upholds, 2 to 1, a designation made on separate grounds.
- 2026.03.03Anthropic is formally designated a supply chain risk, the first U.S. company to be so labeled, and Claude app downloads pass ChatGPT’s for the first time.
- 2026.02.27President Trump orders federal agencies to stop using Anthropic; the same day, OpenAI announces a contract for the Pentagon’s classified network.
- 2026.02.24The Pentagon tells Anthropic to drop its conditions by February 27, and Anthropic refuses.
Statement on the comments from Secretary of War Pete HegsethPentagon labels AI company Anthropic a supply chain risk ‘effective immediately’Anthropic loses legal fight to shed DOD’s designation as a ‘supply-chain risk’
Judge temporarily blocks Trump administration’s Anthropic banAnthropic loses legal fight to shed DOD’s designation as a ‘supply-chain risk’
Pentagon labels AI company Anthropic a supply chain risk ‘effective immediately’
OpenAI reaches Pentagon agreement as Trump orders Anthropic off federal systems
Statement on the comments from Secretary of War Pete HegsethHegseth demands full military access to Anthropic’s AI model Claude and sets deadline for end of week
2026.02.26 · Jobs
Block Cuts 40% of Its Staff
A company with strong results let 40% of its staff go, citing AI, and its stock rose. Research also suggested that entry-level jobs were the first to shrink.
- 2026.02.26Block announces that it will cut more than 4,000 of its roughly 10,000 employees, and Jack Dorsey says the business is strong and AI tools have changed how the company runs.
- 2026.02.26Block shares rise more than 25% in after-hours trading right after the announcement.
- 2025.10.28Amazon cuts about 14,000 corporate jobs, saying it will use AI to run the company leaner.
- 2025.08.26Stanford researchers report that employment of 22- to 25-year-olds in the occupations most exposed to AI fell 13% in relative terms, while employment of experienced workers held steady.
Block cuts 4,000 jobs as Jack Dorsey cites AICanaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence
Block cuts 4,000 jobs as Jack Dorsey cites AI
Block cuts 4,000 jobs as Jack Dorsey cites AIBlock shares surge as Jack Dorsey announces AI-driven layoffs of 4,000 workers
Amazon to cut 14,000 corporate jobs amid AI-driven restructuring
Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence
2026.02.09 · Ads
Ads Come to ChatGPT
ChatGPT, with hundreds of millions of weekly users, began to make money from ads. A rival made its lack of ads the point of a Super Bowl commercial.
- 2026.02.09Ads begin to appear in ChatGPT for some U.S. users, with none shown to minors or in conversations about health or politics.
- 2026.02.04Anthropic pledges to keep ads out of Claude and releases a Super Bowl commercial saying “Ads are coming to AI. But not to Claude.”
- 2026.01.16OpenAI announces that it will test ads on its Free and Go plans, shown below answers and clearly labeled.
OpenAI starts testing ads in ChatGPTAnthropic Promises Claude Will Remain Ad-Free, Mocks ChatGPT Ads in Super Bowl Commercial
OpenAI starts testing ads in ChatGPT
Anthropic Promises Claude Will Remain Ad-Free, Mocks ChatGPT Ads in Super Bowl Commercial
Our approach to advertising and expanding access to ChatGPTOpenAI To Test ChatGPT Ads At The Bottom Of Answers
2026.02.03 · Markets
The SaaSpocalypse
Fear that AI agents could replace the seats in software subscriptions sent software stocks down all at once.
- 2026.02.05Claude Opus 4.6 and GPT-5.3-Codex come out on the same day.
- 2026.02.03Software, legal information and financial data stocks plunge, erasing about $285 billion in a day.
- 2026.02.02SpaceX acquires xAI, for a combined valuation of $1.25 trillion.
- 2026.01.30Anthropic publishes 11 Claude Cowork plugins for work such as legal, sales and finance on GitHub.
anthropics/knowledge-work-pluginsAnthropic’s Claude plugins spark $285 billion software stock selloff as AI targets entire SaaS workflowsMusk’s xAI, SpaceX combo is the biggest merger of all time, valued at $1.25 trillion
Claude Opus 4.6Introducing GPT-5.3-Codex
Anthropic’s Claude plugins spark $285 billion software stock selloff as AI targets entire SaaS workflows
Musk’s xAI, SpaceX combo is the biggest merger of all time, valued at $1.25 trillion
anthropics/knowledge-work-pluginsAnthropic’s Legal Plugin for Claude Cowork May Be the Opening Salvo In A Competition Between Foundation Models and Legal Tech Incumbents
2026.01.29 · World Models · Added 2026.11.07
Models That Build Worlds
World models, which generate explorable environments in real time from text, moved beyond the lab. Google opened a prototype to subscribers while calling it experimental, and AMI Labs raised seed money to build world models as an alternative to large language models.
- 2026.10.27Google DeepMind releases Genie 4 as a research preview, its first real-time world model to generate audio natively, and says details and object permanence can drift over longer sessions.
- 2026.03.10AMI Labs, co-founded by Yann LeCun, raises a $1.03 billion seed round to build world models, at a reported valuation of $3.5 billion.
- 2026.01.29Google opens the experimental prototype Project Genie to Google AI Ultra subscribers in the U.S., and the next day Unity closes down 24% while Take-Two and Roblox also fall.
- 2025.08.05Google DeepMind unveils Genie 3, a world model that generates interactive environments from text in real time at 720p and 24 frames per second, as a limited research preview.
Project Genie: Experimenting with infinite, interactive worldsGenie 4: A new generation of world simulation
Genie 4: A new generation of world simulation
Yann LeCun’s Paris-based AI world model startup raises more than $1BNAMI Labs: Real World. Real Intelligence.
Project Genie: Experimenting with infinite, interactive worldsUnity, Video Game Stocks Fall as Google AI Tool Sparks FearsVideogame stocks slide on Google’s AI model that turns prompts into playable worlds
Genie 3: A new frontier for world models
2026.01.22 · Law and Agents
The AI Basic Act, and a Social Network for Agents
Korea became the first country in the world to put a comprehensive AI law fully into effect. The same month, an open-source personal agent spread rapidly.
- 2026.01.30The personal agent Clawdbot settles on the name OpenClaw after two renames and passes 100,000 stars on GitHub.
- 2026.01.28Moltbook, a social network where only AI agents post, opens, and hundreds of thousands of agents join within days.
- 2026.01.22Korea’s AI Basic Act takes effect, requiring labels on generative AI output, with fines deferred for at least a year.
- 2026.01.09xAI restricts Grok’s image generation over sexual deepfakes, and Indonesia and Malaysia then block access.
“규제보다 진흥” AI기본법 시대… 계도기간 주고 컨설팅 지원Introducing OpenClawOpenClaw and Moltbook explained: The latest AI agent craze
Introducing OpenClawOpenClaw and Moltbook explained: The latest AI agent craze
OpenClaw and Moltbook explained: The latest AI agent craze
“규제보다 진흥” AI기본법 시대… 계도기간 주고 컨설팅 지원
Grok to offer image generation only to paid subscribers after backlashMusk’s Grok blocked by Indonesia, Malaysia over sexualized images in world first
2026.01.08 · China
China’s AI Listing Wave
Under U.S. chip controls, Chinese GPU makers listed in Shanghai and saw their shares multiply on day one. Chinese AI labs then listed in Hong Kong, five months before U.S. labs filed to go public.
- 2026.01.09MiniMax’s shares double on their first day of trading in Hong Kong.
- 2026.01.08Zhipu (Z.ai) becomes the first of China’s six “AI tigers” to go public, in Hong Kong, and closes 13% higher on its first day.
- 2025.12.17MetaX, another Chinese GPU maker, closes its first trading day up 693%.
- 2025.12.05Moore Threads, which makes GPUs in China under U.S. export controls, lists on Shanghai’s STAR Market and soars 468% on its first day.
Moore Threads soars 468% in debut as China accelerates its GPU pushChinese AI Chipmaker MetaX Skyrockets in Shanghai Stock Market DebutMiniMax jumps 91% in Hong Kong debut after $619m IPO
MiniMax, China’s second AI tiger to go public, doubles in value in Hong Kong debut
AI large-model company Zhipu gains more than 13% in HK debutStartups go public in litmus test for Chinese AI
Chinese AI Chipmaker MetaX Skyrockets in Shanghai Stock Market Debut
Moore Threads soars 468% in debut as China accelerates its GPU push
2025.12.11 · State Law
The States and Washington
After Congress declined to block state AI laws, California passed a frontier AI safety law. At the end of the year, the president moved against state laws by executive order.
- 2025.12.11President Trump signs an executive order that sets up a task force to challenge state AI laws in court and allows some federal broadband funds to be withheld from states with such laws.
- 2025.09.29A year after vetoing SB 1047, Governor Newsom signs SB 53, requiring frontier AI developers to publish safety frameworks and report critical safety incidents.
- 2025.07.23The White House releases America’s AI Action Plan, and the president signs three executive orders, including ones to speed data center permits and promote exports of American AI technology.
- 2025.07.01The U.S. Senate votes 99 to 1 to strip a 10-year ban on state AI regulation from the budget bill.
Fact Sheet: President Donald J. Trump Ensures a National Policy Framework for Artificial IntelligenceGovernor Newsom signs SB 53, advancing California’s world-leading artificial intelligence industry
Fact Sheet: President Donald J. Trump Ensures a National Policy Framework for Artificial Intelligence
Governor Newsom signs SB 53, advancing California’s world-leading artificial intelligence industry
The Trump Administration’s 2025 AI Action Plan – Winning the Race: America’s AI Action Plan – and Related Executive Orders
Senate Strikes AI Moratorium from Budget Reconciliation Bill in Overwhelming 99-1 Vote
2025.11.25 · Music
Deals Instead of Lawsuits
A song made by AI reached No. 1 on a chart. Record labels that had sued AI music services dropped their suits and signed licensing deals.
- 2025.11.25Warner Music Group also settles with Suno and agrees to build licensed music models together.
- 2025.11.19Warner Music Group settles with Udio; the same day, Suno raises $250 million at a $2.45 billion valuation.
- 2025.11.08AI singer Breaking Rust’s “Walk My Walk” reaches No. 1 on Billboard’s Country Digital Song Sales chart, reported as the first AI song to top a Billboard chart.
- 2025.10.29Universal Music Group settles its lawsuit with Udio and agrees to build a licensed music platform together, while Udio angers users by blocking song downloads.
Universal Music settles Udio lawsuit, strikes deal for licensed AI music platformFake country artist goes to2025 in artificial intelligence
Warner Music Group strikes ‘landmark’ deal with Suno; settles copyright lawsuit against AI music generator
AI Music Generators: Warner Music Group and Udio Settle Copyright Lawsuit as Suno Raises $250M to Fight On
Fake country artist goes to
Universal Music settles Udio lawsuit, strikes deal for licensed AI music platform
2025.11.18 · Overtaking
Gemini 3 and “Code Red”
When Google’s Gemini 3 took the lead on major benchmarks, OpenAI went on alert internally.
- 2025.12.09Anthropic donates MCP to the Agentic AI Foundation under the Linux Foundation.
- 2025.12.01Altman sends a “code red” memo to staff, and GPT-5.2’s release is moved up to December 11.
- 2025.11.24Anthropic releases Claude Opus 4.5.
- 2025.11.18Google releases Gemini 3.
A new era of intelligence with Gemini 3Sam Altman declares ‘Code Red’ as Google’s Gemini surges, three years after ChatGPT caused Google CEO Sundar Pichai to do the sameDonating MCP to the Agentic AI Foundation
Donating MCP to the Agentic AI Foundation
Sam Altman declares ‘Code Red’ as Google’s Gemini surges, three years after ChatGPT caused Google CEO Sundar Pichai to do the same
Introducing Claude Opus 4.5
A new era of intelligence with Gemini 3
2025.11.13 · Cyber
Hacking Run by AI
Coding agents were turned to hacking as well. With only a few approvals from people, AI carried out most of the work, from reconnaissance to intrusion.
- 2025.11.14China’s foreign ministry dismisses the report as an accusation made without evidence and says China opposes hacking.
- 2025.11.13Anthropic discloses an espionage campaign by a group it assesses as Chinese state-sponsored, which used Claude Code against about 30 targets, with AI doing 80 to 90% of the work and a few intrusions succeeding.
- 2025.08.27Anthropic reports that a criminal group used Claude Code to break into at least 17 organizations, including healthcare, emergency services and government agencies, and demanded ransoms, some above $500,000.
Disrupting an AI-orchestrated cyber espionage campaignCrims laud Claude, use Anthropic’s AI to plant ransomwareAnthropic says Chinese hackers used its AI in online attack, a claim refuted by China
Anthropic says Chinese hackers used its AI in online attack, a claim refuted by China
Disrupting an AI-orchestrated cyber espionage campaign
Crims laud Claude, use Anthropic’s AI to plant ransomware
2025.10.31 · Korea
Gyeongju, 260,000 GPUs
In Korea for the APEC summit, NVIDIA CEO Jensen Huang announced that NVIDIA would supply 260,000 of the latest GPUs to the Korean government and companies.
- 2025.10.31The government, Samsung, SK and Hyundai Motor Group get 50,000 each, and NAVER Cloud gets 60,000.
- 2025.10.31The presidential office says Huang told a closed-door meeting that the United States has the most GPUs, followed by China, and that Korea’s roughly 300,000 will rank third.
- 2025.10.30At a fried chicken restaurant in Samseong-dong, Seoul, Huang meets Samsung Electronics Executive Chairman Jay Y. Lee and Hyundai Motor Group Executive Chair Euisun Chung over chicken and beer.
2026.11.02 Corrected “Huang says Korea will have the most GPUs of any country after the United States and China.” According to reporting, Huang named the United States and China in a closed-door meeting, and the ranking of Korea third came from the presidential office (The Korea Herald, How 260,000 Nvidia chips could redefine Korea’s AI ambitions).
NVIDIA, South Korea Government and Industrial Giants Build AI Infrastructure and Ecosystem to Fuel Korea Innovation, Industries and Jobs젠슨 황 덕에 한숨 돌린 정부…“AI 적기투자 필요, 예산증액·추경도”
NVIDIA, South Korea Government and Industrial Giants Build AI Infrastructure and Ecosystem to Fuel Korea Innovation, Industries and Jobs
How 260,000 Nvidia chips could redefine Korea’s AI ambitions
Jensen Huang a rock star as Nvidia, Samsung, Hyundai chiefs celebrate partnership over chicken and beer
2025.10.28 · Governance
A Nonprofit in Charge of a For-Profit
OpenAI, which began as a nonprofit lab, finished turning its for-profit arm into a public benefit corporation. Its nonprofit foundation kept control, and its deal with Microsoft was rewritten.
- 2025.10.28OpenAI completes its recapitalization: the for-profit becomes OpenAI Group PBC, the nonprofit OpenAI Foundation holds about 26%, worth about $130 billion, and the attorneys general of California and Delaware agree not to oppose it.
- 2025.10.28Microsoft signs a new agreement giving it about 27%, worth about $135 billion, under which any OpenAI declaration of AGI will be verified by an independent expert panel and OpenAI commits to $250 billion more in Azure services.
- 2025.05.05OpenAI says its for-profit arm will become a public benefit corporation but stay under the nonprofit’s control, dropping a plan to loosen that control.
OpenAI caves to pressure, keeps nonprofit in chargeOpenAI simplifies corporate structure for easier investmentOpenAI completes for-profit restructuring and grants Microsoft a 27% stake
OpenAI simplifies corporate structure for easier investment
OpenAI completes for-profit restructuring and grants Microsoft a 27% stake
OpenAI caves to pressure, keeps nonprofit in charge
2025.09.22 · Compute
Deals by the Gigawatt
The race for models spread into a race for power and chips. Data center deals were signed by the gigawatt.
- 2025.10.29NVIDIA becomes the first company in the world worth $5 trillion.
- 2025.09.30OpenAI launches Sora 2 and Sora, a social video app.
- 2025.09.22NVIDIA announces its intent to invest up to $100 billion in OpenAI (10 GW), and around the same time, OpenAI also signs large deals with Oracle, AMD and Broadcom.
- 2025.09.05Anthropic agrees to a $1.5 billion settlement in an authors’ class action, the largest ever in an AI copyright dispute.
OpenAI and NVIDIA Announce Strategic Partnership to Deploy 10 Gigawatts of NVIDIA SystemsAI company Anthropic agrees to pay $1.5B to settle lawsuit with authorsSora 2 is here
Nvidia Becomes the First Company to Reach a $5 Trillion Market Value, Leaving Apple and Microsoft in the Dust
Sora 2 is here
OpenAI and NVIDIA Announce Strategic Partnership to Deploy 10 Gigawatts of NVIDIA Systems
AI company Anthropic agrees to pay $1.5B to settle lawsuit with authors
2025.09.17 · Chip Controls
Tug of War Over Chips
The U.S. blocked NVIDIA’s China-market chip, then reopened sales in exchange for a cut of the revenue. This time China told its companies not to buy NVIDIA chips and turned to its own.
- 2025.09.18Huawei lays out its Ascend roadmap, with a new chip each year starting with the Ascend 950 in 2026, its own HBM and SuperPoDs linking 8,192 chips.
- 2025.09.17China’s Cyberspace Administration tells ByteDance, Alibaba and others to stop testing and ordering NVIDIA’s China-market RTX Pro 6000D, the Financial Times reports.
- 2025.08.10NVIDIA and AMD agree to pay the U.S. government 15% of their revenue from H20 and MI308 sales in China as export licenses begin to be issued, a U.S. official says.
- 2025.04.15NVIDIA discloses that exports of its China-market H20 now need a license for the indefinite future and warns of up to $5.5 billion in charges.
NVIDIA Corporation Form 8-K (April 2025)Nvidia, AMD to pay 15% of China chip sale revenues to US, official saysChina tells tech firms to stop buying Nvidia’s AI chips, FT reports
Huawei outlines roadmap for Ascend AI chips
China tells tech firms to stop buying Nvidia’s AI chips, FT reports
Nvidia, AMD to pay 15% of China chip sale revenues to US, official says
NVIDIA Corporation Form 8-K (April 2025)
2025.08.26 · Comfort
The Children Who Confided in AI
The death of a teenager who had been pouring his heart out to a chatbot went to court. Companies rushed out protections for minors.
- 2025.10.29Character.AI announces that from November 25 it will block open-ended chat for users under 18.
- 2025.09.29OpenAI launches parental controls that let parents link and manage their children’s accounts.
- 2025.09.11The U.S. Federal Trade Commission orders seven companies, including OpenAI, Meta and Character.AI, to report on their safety measures for chatbots used by teenagers.
- 2025.08.26The parents of 16-year-old Adam Raine sue OpenAI and Sam Altman, citing their son’s conversations with ChatGPT.
2025 in artificial intelligenceCharacter.AI is ending its chatbot experience for kids
Character.AI is ending its chatbot experience for kids
Introducing parental controls
FTC Launches Inquiry into AI Chatbots Acting as Companions
Parents of 16-year-old sue OpenAI, claiming ChatGPT advised on teen’s suicide
2025.08.07 · Summer
The Talent War and GPT-5
Meta spent heavily to lure researchers, and AI posted gold-medal scores at the math olympiad. In August, OpenAI released GPT-5 and gpt-oss, its first open-weight model in six years.
- 2025.08.26Google releases Nano Banana (Gemini 2.5 Flash Image), setting off a photo-editing craze.
- 2025.08.07GPT-5 launches, and OpenAI, which had removed its earlier models, brings GPT-4o back after users push back.
- 2025.07.19An experimental OpenAI model and Google’s Gemini Deep Think reach gold-medal level at the International Mathematical Olympiad.
- 2025.06.30Meta puts $14.3 billion into Scale AI, hires Alexandr Wang and launches a superintelligence lab.
Introducing GPT-5Introducing gpt-ossAdvanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad
Introducing Gemini 2.5 Flash Image, our state-of-the-art image model
Introducing GPT-5‘We hear you’: Sam Altman restores GPT-4o for plus subscribers after GPT-5 backlash
Advanced version of Gemini with Deep Think officially achieves gold-medal standard at the International Mathematical Olympiad
Meta ramps up AI efforts by paying $14.3 billion for 49% of Scale and hiring its CEO for ‘superintelligence’ teamMark Zuckerberg overhauled Meta’s entire AI org in a risky, multibillion-dollar bet on ‘superintelligence’
2025.07.11 · Open Models
China Opens Its Weights
After the DeepSeek shock, Alibaba and Moonshot AI released one open-weight model after another, and Chinese models began to contend for the top spot among open models.
- 2025.11.06Moonshot AI releases Kimi K2 Thinking, which it says beats GPT-5 and Claude Sonnet 4.5 on Humanity’s Last Exam with tools and can chain 200 to 300 tool calls.
- 2025.07.11Moonshot AI releases the weights of Kimi K2, a one-trillion-parameter model built for agentic tool use.
- 2025.04.29Alibaba releases Qwen3, ranging from six dense models under Apache 2.0 to a 235-billion-parameter mixture-of-experts model, covering 119 languages and dialects.
Qwen3: Think Deeper, Act Fastermoonshotai/Kimi-K2-Instructmoonshotai/Kimi-K2-Thinking
Moonshot launches open-source ‘Kimi K2 Thinking’ AI with a trillion parameters and reasoning capabilities
moonshotai/Kimi-K2-Instruct
Qwen3: Think Deeper, Act Faster
2025.06.23 · Copyright
Is Training Fair Use?
U.S. courts issued their first rulings on whether training AI on copyrighted works is fair use. The judges weighed the question differently.
- 2026.09.29The Third Circuit upholds the ROSS ruling, the first federal appeals decision on fair use in AI training.
- 2025.06.25Judge Vince Chhabria rules for Meta, but only because the authors failed to show market harm, noting that an argument that AI floods the market with competing works could win.
- 2025.06.23Judge William Alsup rules that Anthropic’s training of Claude on books was fair use, but that keeping pirated copies in a central library was not.
- 2025.02.11A federal judge rules that ROSS Intelligence’s copying of 2,243 Westlaw headnotes to build an AI legal search tool was not fair use, in a case that did not involve generative AI.
Two Northern District of California Judges Find Some Copying to Train Generative AI Is Fair UseMeta Wins AI Copyright Case, But Judge Writes Roadmap for Authors’ Revenge
Third Circuit Rejects Fair Use for ROSS’s Copying of Westlaw HeadnotesThird Circuit Affirms Revised Fair Use Ruling Against ROSS’ AI Legal Research Platform in Sealed Opinion
Meta Wins AI Copyright Case, But Judge Writes Roadmap for Authors’ RevengeTwo Northern District of California Judges Find Some Copying to Train Generative AI Is Fair Use
Two Northern District of California Judges Find Some Copying to Train Generative AI Is Fair Use
Revised Fair Use Ruling Finds No Transformative Use in Developing AI Search Tool
2025.05.22 · Agents
The Age of Coding Agents
Within a week, three companies released coding agents and new models. AI began to do hours-long development work on its own.
- 2025.05.22Anthropic releases Claude Opus 4 and Claude Sonnet 4, and Claude Code becomes generally available.
- 2025.05.21OpenAI acquires io, Jony Ive’s hardware startup, for about $6.5 billion.
- 2025.05.20At Google I/O, Google unveils Veo 3, a video model that makes sound as well, and Jules, a coding agent.
- 2025.05.16OpenAI releases Codex, a coding agent that works in the cloud.
Introducing CodexGoogle I/O 2025: From research to realityIntroducing Claude 4
Introducing Claude 4
OpenAI to Buy AI Device Startup From Apple Veteran Jony Ive in $6.5 Billion Deal
Google I/O 2025: From research to reality
Introducing Codex
2025.05.13 · The Gulf
Chips Head to the Gulf
Once Washington dropped a rule that would have rationed AI chips by country, large chip deals for Saudi Arabia and the UAE followed during the president’s tour of the Middle East.
- 2025.05.22G42, OpenAI, Oracle, NVIDIA, SoftBank and Cisco announce Stargate UAE in Abu Dhabi, a 1 GW cluster within a planned 5 GW AI campus, with the first 200 MW due in 2026.
- 2025.05.13The Commerce Department rescinds the Biden administration’s AI Diffusion Rule, which would have sorted countries into tiers for AI chip exports, just before it was to take effect, and warns that using Huawei Ascend chips may violate U.S. export controls.
- 2025.05.13During President Trump’s visit to Riyadh, NVIDIA agrees to supply HUMAIN, owned by Saudi Arabia’s Public Investment Fund, starting with an 18,000-chip GB300 supercomputer and adding several hundred thousand GPUs over five years.
Department of Commerce Announces Rescission of Biden-Era Artificial Intelligence Diffusion Rule, Strengthens Chip-Related Export ControlsSaudi Arabia and NVIDIA to Build AI Factories to Power Next Wave of Intelligence for the Age of ReasoningGlobal Tech Alliance Launches Stargate UAE
Global Tech Alliance Launches Stargate UAE
Department of Commerce Announces Rescission of Biden-Era Artificial Intelligence Diffusion Rule, Strengthens Chip-Related Export Controls
Nvidia, AWS and AMD announce Humain AI partnerships on Trump’s Saudi visit
2025.03.25 · Mainstream
The Ghibli Craze
Once ChatGPT gained new image generation, turning photos into Ghibli-style pictures became a pastime that spread around the world.
- 2025.03.31OpenAI raises $40 billion at a $300 billion valuation, and ChatGPT reaches 500 million weekly users.
- 2025.03.25Image generation in GPT-4o is released, and Altman says demand is so heavy that the GPUs are melting.
- 2025.03.25The same day, Google releases Gemini 2.5 Pro.
- 2025.03.06Manus, a general-purpose AI agent, is unveiled, and people scramble for invitation codes.
Introducing 4o Image GenerationGemini 2.5: Our most intelligent AI modelNew funding to build towards AGI
New funding to build towards AGIOpenAI secures $40 billion in record-breaking funding round, valuation hits $300 billion
Introducing 4o Image GenerationSam Altman says ChatGPT’s viral image-generation AI is ‘melting’ OpenAI’s GPUs
Gemini 2.5: Our most intelligent AI model
Manus AI launched in China, challenging GPT and DeepSeekPeople are listing invite codes to Chinese AI agent Manus for $1,000 and more on resale sites
2025.02.24 · Coding
Vibe Coding
A way of building software by telling AI what to do, without reading the code, got a name. A coding agent that works in the terminal arrived.
- 2025.02.24Anthropic releases Claude 3.7 Sonnet and the coding agent Claude Code as a research preview.
- 2025.02.11At the AI Action Summit in Paris, the United States and the United Kingdom do not sign the joint declaration.
- 2025.02.02Andrej Karpathy coins the term “vibe coding.”
- 2025.02.02OpenAI releases deep research, which searches the web and writes reports.
There’s a new kind of coding I call “vibe coding”Introducing deep researchClaude 3.7 Sonnet and Claude Code
Claude 3.7 Sonnet and Claude Code
As US and UK refuse to sign the Paris AI Action Summit statement, other countries commit to developing ‘open, inclusive, ethical’ AI
There’s a new kind of coding I call “vibe coding”
Introducing deep research
2025.02.06 · Reasoning Models · Added 2026.11.07
Whether to Show the Thinking
Step-by-step reasoning began as a prompting technique from Google researchers in 2022. OpenAI trained it into o1 with reinforcement learning and hid the raw reasoning. As Google, DeepSeek and Anthropic released models that show their thinking, OpenAI opened up part of it.
- 2025.02.24Anthropic lets Claude 3.7 Sonnet users pick near-instant answers or extended thinking that they can read, and API users can set how long the model thinks.
- 2025.02.06After DeepSeek’s R1 shows its full reasoning, OpenAI makes o3-mini display more of its thought process, in a post-processed form rather than the raw chain of thought.
- 2024.12.19Google releases Gemini 2.0 Flash Thinking, an experimental model that shows its thoughts, in public preview.
- 2024.09.16OpenAI is reported to be emailing users who probe o1’s hidden reasoning, warning that further violations may cost them access.
Learning to reason with LLMsChain-of-Thought Prompting Elicits Reasoning in Large Language Models
Claude 3.7 Sonnet and Claude Code
OpenAI now reveals more of its o3-mini model’s thought process
Gemini API changelogGoogle releases Gemini 2.0 Flash Thinking, an AI model that introduces a thought process to enhance inference, and exceeds OpenAI’s o1-preview and GPT-4o in various tests
Ban warnings fly as users dare to probe the “thoughts” of OpenAI’s latest model
2025.01.27 · Shock
The DeepSeek Shock
A small Chinese lab released an o1-class reasoning model, weights and all. U.S. stock markets were shaken.
- 2025.01.27The DeepSeek app tops the U.S. App Store, and NVIDIA shares fall 17% in a day, erasing about $589 billion in market value.
- 2025.01.23OpenAI releases Operator, an agent that works a web browser on the user’s behalf.
- 2025.01.21President Trump announces Stargate, a $500 billion data center venture of OpenAI, SoftBank and Oracle.
- 2025.01.20DeepSeek releases its reasoning model R1 under the MIT license.
DeepSeek-R1 ReleaseAnnouncing The Stargate ProjectNvidia sheds almost $600 billion in market cap, biggest one-day loss in U.S. history
Nvidia sheds almost $600 billion in market cap, biggest one-day loss in U.S. historyDeepSeek AI assistant surpasses ChatGPT on US App Store
Introducing Operator
Announcing The Stargate Project
DeepSeek-R1 Release
2024.11.25 · The Agent’s Hands
AI That Sees the Screen, and o3
AI began to look at the screen, move the mouse and connect to outside tools through a standard protocol. At the end of the year, reasoning models took another leap.
- 2024.12.26DeepSeek releases V3, claiming a training cost of about $5.6 million; the same day, Korea’s National Assembly passes the AI Basic Act.
- 2024.12.20OpenAI announces o3, which scores up to 87.5% on ARC-AGI, a test of abstract reasoning.
- 2024.11.25Anthropic releases MCP, an open standard that connects AI to data and tools.
- 2024.10.22Anthropic launches a beta of computer use, a feature that looks at the screen and works the mouse and keyboard.
Computer use, Claude 3.5 Sonnet, and Claude 3.5 HaikuIntroducing the Model Context ProtocolOpenAI o3 Breakthrough High Score on ARC-AGI-Pub
Introducing DeepSeek-V3DeepSeek-V3 Technical ReportSouth Korea National Assembly Passes New AI Bill
OpenAI o3 Breakthrough High Score on ARC-AGI-Pub
Introducing the Model Context Protocol
Computer use, Claude 3.5 Sonnet, and Claude 3.5 Haiku
2024.09.29 · State Law
The SB 1047 Veto
A California bill that would have required developers of large AI models to put safeguards against catastrophic harm in place split the industry. The governor vetoed it, and the fight over regulating frontier AI at the state level went on.
- 2024.09.29Governor Newsom vetoes the bill, saying that targeting only the largest, most expensive models could give a false sense of security and ignores whether a system is deployed in high-risk settings.
- 2024.08.28The California Assembly passes the bill, and the Senate gives final approval the next day, sending it to Governor Gavin Newsom.
- 2024.08.21OpenAI writes to oppose SB 1047, saying it would slow innovation, while Anthropic writes that the amended bill’s benefits likely outweigh its costs.
SB 1047 Veto MessageGov. Newsom vetoes California’s controversial AI bill, SB 1047
AI Public Safety Risk Legislation Sent to California Governor
OpenAI Says California’s Controversial AI Bill Will Hurt InnovationAnthropic says California AI bill’s benefits likely outweigh costs
2024.09.20 · Power
Data Centers Go Nuclear
As AI data centers drew more power, Big Tech began signing deals directly for nuclear energy. One deal was set to restart a reactor at Three Mile Island, the site of the 1979 accident.
- 2024.10.16Amazon invests in small modular reactor developer X-energy and sets a goal of bringing more than 5 GW of nuclear power online by 2039.
- 2024.10.14Google signs a deal with Kairos Power to buy 500 MW from a fleet of small modular reactors by 2035, with the first planned for 2030.
- 2024.09.20Constellation signs a 20-year power purchase agreement with Microsoft to restart Three Mile Island Unit 1, shut in 2019, and bring about 835 MW back online by 2028.
Constellation to Launch Crane Clean Energy Center, Restoring Jobs and Carbon-Free Power to the GridGoogle and Kairos Power Partner to Deploy 500 MW of Clean Electricity GenerationAmazon announces small modular reactor deals with Dominion, X-energy, Energy Northwest
Amazon announces small modular reactor deals with Dominion, X-energy, Energy Northwest
Google and Kairos Power Partner to Deploy 500 MW of Clean Electricity Generation
Constellation to Launch Crane Clean Energy Center, Restoring Jobs and Carbon-Free Power to the Grid
2024.09.12 · Reasoning
Models That Think, and Two Nobel Prizes
A reasoning model trained to think at length before answering arrived. A month later, two Nobel Prizes went to AI research.
- 2024.10.09The Nobel Prize in Chemistry goes to Demis Hassabis and John Jumper for AlphaFold and to David Baker for protein design.
- 2024.10.08The Nobel Prize in Physics goes to John Hopfield and Geoffrey Hinton, who laid the foundations of artificial neural networks.
- 2024.10.02OpenAI raises $6.6 billion at a valuation of $157 billion.
- 2024.09.12OpenAI releases a preview of its reasoning model o1, which pulls far ahead on math and coding problems.
Introducing OpenAI o1-previewPress release: The Nobel Prize in Physics 2024Press release: The Nobel Prize in Chemistry 2024
Press release: The Nobel Prize in Chemistry 2024
Press release: The Nobel Prize in Physics 2024
OpenAI has raised $6.6 billion in a mega-funding deal that values the ChatGPT maker at $157 billion
Introducing OpenAI o1-preview
2024.06.18 · Where the Money Went
NVIDIA, Number One in the World
As demand for AI chips piled up, NVIDIA became the world’s most valuable company for the first time. Apple said it would bring generative AI to the iPhone.
- 2024.06.20Anthropic releases Claude 3.5 Sonnet, along with Artifacts, which shows results in a window beside the chat.
- 2024.06.19Sutskever founds SSI, which says it will build only safe superintelligence.
- 2024.06.18NVIDIA’s market value reaches about $3.3 trillion, passing Microsoft to rank first for the first time.
- 2024.06.10Apple announces Apple Intelligence, and Siri passes hard questions to ChatGPT.
Nvidia topples Microsoft as world’s most valuable companyIntroducing Apple Intelligence for iPhone, iPad, and MacIntroducing Claude 3.5 Sonnet
Introducing Claude 3.5 Sonnet
OpenAI co-founder Ilya Sutskever launches startup to develop ‘safe superintelligence’
Nvidia topples Microsoft as world’s most valuable company
Introducing Apple Intelligence for iPhone, iPad, and Mac
2024.05.21 · China
The Price War
When DeepSeek and ByteDance offered models at rock-bottom prices, Alibaba and Baidu cut theirs within days. Eight months later, the same DeepSeek’s R1 shook U.S. markets.
- 2024.05.21Alibaba Cloud cuts prices on its models by up to 97%.
- 2024.05.21The same day, Baidu makes ERNIE Lite and ERNIE Speed free.
- 2024.05.15ByteDance launches its Doubao models for businesses at 0.0008 yuan per thousand tokens, which it says is 99.3% below the industry average.
- 2024.05.06DeepSeek releases DeepSeek-V2 with open weights and prices its API at 1 yuan per million input tokens and 2 yuan per million output tokens.
ByteDance surprises AI rivals with ultra-low cost Doubao modelPrice war in China’s AI field expanded as Alibaba Cloud and Baidu have drastically reduced model prices
Price war in China’s AI field expanded as Alibaba Cloud and Baidu have drastically reduced model prices
Price war in China’s AI field expanded as Alibaba Cloud and Baidu have drastically reduced model prices
ByteDance’s Volcano Engine Starts Price War in China’s Enterprise LLMs Sector
DeepSeek-V2AI“快进”到价格战阶段？从Deepseek-V2定价看大模型降本新风向
2024.05.13 · Voice
GPT-4o Hears, Speaks and Sees
A voice AI was demonstrated that could be interrupted mid-sentence and could laugh as it talked. Seoul hosted the second AI summit.
- 2024.05.21At the AI Seoul Summit, 16 companies sign the Frontier AI Safety Commitments.
- 2024.05.14At Google I/O, Google puts AI summaries (AI Overviews) at the top of search results.
- 2024.05.14Ilya Sutskever leaves OpenAI, and the Superalignment team in charge of long-term safety is disbanded.
- 2024.05.13OpenAI releases GPT-4o for real-time voice conversation, then pulls one voice after Scarlett Johansson objects that it sounds like hers.
Hello GPT-4oHow the voices for ChatGPT were chosenFrontier AI Safety Commitments, AI Seoul Summit 2024
Frontier AI Safety Commitments, AI Seoul Summit 2024
Generative AI in Search: Let Google do the searching for you
OpenAI Dissolves High-Profile Safety Team After Chief Scientist Sutskever’s Exit
Hello GPT-4oHow the voices for ChatGPT were chosen
2024.03.04 · Rivalry
Claude 3 and the “AI Developer”
Claude 3 matched or beat GPT-4 on several benchmarks. A demo of an agent that writes code on its own unsettled developers.
- 2024.03.18NVIDIA announces Blackwell, its next-generation GPU.
- 2024.03.13The European Parliament passes the AI Act, the world’s first comprehensive law regulating AI.
- 2024.03.12Cognition demonstrates Devin, an “AI software engineer.”
- 2024.03.04Anthropic releases the Claude 3 model family: Opus, Sonnet and Haiku.
Introducing the next generation of ClaudeIntroducing Devin, the first AI software engineerArtificial Intelligence Act: MEPs adopt landmark law
NVIDIA Blackwell Platform Arrives to Power a New Era of Computing
Artificial Intelligence Act: MEPs adopt landmark law
Introducing Devin, the first AI software engineer
Introducing the next generation of Claude
2024.02.15 · Video
Sora Turns Sentences into Video
OpenAI demonstrated Sora, which makes realistic videos up to a minute long from text alone. The same day, Google released a model that reads a million tokens.
- 2024.02.22Google pauses Gemini’s image generation of people after criticism that it depicted historical figures inaccurately.
- 2024.02.15OpenAI demonstrates Sora; the same day, Google announces Gemini 1.5, with a context of one million tokens.
- 2023.12.27The New York Times sues OpenAI and Microsoft, saying they trained on its articles without permission.
- 2023.12.06Google announces Gemini 1.0, a multimodal model, and the following February renames Bard to Gemini as well.
Sora: Creating video from textOur next-generation model: Gemini 1.5Complaint, The New York Times Company v. Microsoft Corporation et al., Case 1:23-cv-11195 (S.D.N.Y. Dec. 27, 2023)
Google pauses Gemini AI image generator after it created inaccurate historical pictures
Sora: Creating video from textOur next-generation model: Gemini 1.5
Complaint, The New York Times Company v. Microsoft Corporation et al., Case 1:23-cv-11195 (S.D.N.Y. Dec. 27, 2023)
Introducing Gemini: our largest and most capable AI modelBard becomes Gemini: Try Ultra 1.0 and a new mobile app today
2024.01.21 · Elections
A Fake Voice in an Election Year
A year of elections around the world began with a fake voice made by AI. Within weeks, regulators and companies moved to respond.
- 2024.02.16At the Munich Security Conference, 20 companies announce an accord to combat deceptive AI content in elections.
- 2024.02.08The FCC rules that robocalls using AI-generated voices are illegal under the Telephone Consumer Protection Act.
- 2024.01.28As sexually explicit deepfakes of Taylor Swift spread, X temporarily blocks searches for “Taylor Swift.”
- 2024.01.21Two days before the New Hampshire primary, robocalls in an AI imitation of President Biden’s voice tell voters to “save your vote for the November election.”
FCC votes unanimously to ban unsolicited AI robocallsTechnology industry to combat deceptive use of AI in 2024 electionsAI Elections Accord
Technology industry to combat deceptive use of AI in 2024 electionsAI Elections Accord
FCC votes unanimously to ban unsolicited AI robocalls
X restores Taylor Swift searches after deepfake explicit images triggered temporary block
New Hampshire robocalls stir up new firestorm over artificial intelligence
2023.11.17 · Rules and Turmoil
The Firing, and Five Days
In an autumn when countries began writing rules for AI, OpenAI’s board fired Sam Altman and reversed course five days later.
- 2023.11.21Altman’s return is agreed, and a new board is formed.
- 2023.11.17OpenAI’s board removes Altman, and most employees sign a letter warning that they will resign en masse.
- 2023.11.01The first AI Safety Summit is held at Bletchley Park in the United Kingdom, where 28 countries and the EU sign a joint declaration.
- 2023.10.30President Biden signs an executive order on AI that requires developers of large models to share safety test results.
OpenAI announces leadership transitionSam Altman returns as CEO, OpenAI has a new initial boardThe Bletchley Declaration by Countries Attending the AI Safety Summit, 1-2 November 2023
Sam Altman returns as CEO, OpenAI has a new initial board
OpenAI announces leadership transition
The Bletchley Declaration by Countries Attending the AI Safety Summit, 1-2 November 2023
FACT SHEET: President Biden Issues Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence
2023.10.17 · Chip Controls
Closing the Workaround
In October 2022, the U.S. barred exports of advanced AI chips to China. When NVIDIA answered with toned-down chips made for China, the U.S. blocked those too a year later.
- 2024.12.02The U.S. puts export controls on high-bandwidth memory (HBM) and chipmaking equipment and adds 140 entities to its Entity List.
- 2023.10.23The U.S. government tells NVIDIA the new rules apply immediately, cutting off the A800 and H800 it had made for the Chinese market.
- 2023.10.17The Commerce Department revises the AI chip export controls it set in October 2022, adding a performance-density test to catch workaround chips and extending them to other arms-embargoed countries and Macau.
Commerce Strengthens Restrictions on Advanced Computing Semiconductors, Semiconductor Manufacturing Equipment, and Supercomputing Items to Countries of ConcernBIS updated public information page on export controls imposed on advanced computing and semiconductor manufacturing items to the People’s Republic of China (PRC)
US Tightens Curbs on China’s Access to AI Memory, Chip Tools
NVIDIA Corporation Form 8-K (October 2023)
2023.08.31 · China
Approved Chatbots
In China, where ChatGPT is not available, only chatbots that followed government rules and won approval could open to the public. The first approvals came two weeks after the rules took effect.
- 2023.08.31Baidu, ByteDance, SenseTime and others receive the first approvals and open their chatbots to the public.
- 2023.08.31Within 12 hours of opening to the public, ERNIE Bot tops the free-app chart on Apple’s App Store in China.
- 2023.08.15China’s Interim Measures for the Administration of Generative AI Services, issued by seven agencies, take effect, requiring services to uphold core socialist values and those able to sway public opinion to pass a security assessment.
- 2023.03.16Baidu unveils its chatbot ERNIE Bot, opening it only to business partners and people on a waitlist.
生成式人工智能服务管理暂行办法Baidu gets China’s green light to release its ChatGPT-like Ernie Bot to the public11 Chinese tech companies granted permission to fully release ChatGPT-like tools
11 Chinese tech companies granted permission to fully release ChatGPT-like tools
11 Chinese tech companies granted permission to fully release ChatGPT-like toolsBaidu gets China’s green light to release its ChatGPT-like Ernie Bot to the public
生成式人工智能服务管理暂行办法
Baidu launches ChatGPT rival ERNIE Bot, sees 30,000 companies apply to use it in first hour
2023.07.18 · Open Models
Llama 2 and HyperCLOVA X
Meta released a model that could be used commercially, and the open-model camp grew. Korea produced a hyperscale model of its own.
- 2023.08.24NAVER unveils HyperCLOVA X, a model specialized for Korean.
- 2023.07.18Meta releases Llama 2, open to commercial use, together with Microsoft.
- 2023.07.12Elon Musk announces the founding of the AI company xAI.
- 2023.07.11Anthropic releases Claude 2, which reads 100,000 tokens, about a book’s worth, in one go.
Meta and Microsoft Introduce the Next Generation of LlamaClaude 2네이버, ‘하이퍼클로바X’ 공개... “모두를 위한 AI 시대” 활짝
네이버, ‘하이퍼클로바X’ 공개... “모두를 위한 AI 시대” 활짝
Meta and Microsoft Introduce the Next Generation of Llama
Elon Musk Announces New Company xAI as He Seeks to Build ChatGPT Alternative
Claude 2
2023.05.16 · Warning
Warnings from the Builders
A researcher who pioneered deep learning and the heads of AI companies began to speak about the risks themselves. Regulation became a political issue.
- 2023.05.30Hundreds sign a one-sentence statement saying the risk of extinction from AI should be treated like pandemics and nuclear war.
- 2023.05.16Sam Altman proposes a government licensing regime for powerful models at a U.S. Senate hearing.
- 2023.05.02The Writers Guild of America goes on strike, with limits on AI use as a core issue (settled in September).
- 2023.05.01Geoffrey Hinton leaves Google after ten years, saying he wants to speak freely about the risks of AI.
Geoffrey Hinton tells us why he’s now scared of the tech he helped buildTestimony before the U.S. SenateStatement on AI Extinction Risk
Statement on AI Extinction Risk
Testimony before the U.S. Senate
Writers Strike 2023: Hollywood screenwriters don’t want AI taking their jobs eitherHollywood writers strike ends as WGA sends contract to member vote
Geoffrey Hinton tells us why he’s now scared of the tech he helped build
2023.03.14 · Leap in Capability
GPT-4
GPT-4 arrived, reading images and scoring in the top 10% on a simulated U.S. bar exam. The same day, Anthropic released its first Claude.
- 2023.03.31Italy’s data protection authority blocks access to ChatGPT; the block is lifted about a month later.
- 2023.03.22Elon Musk, Steve Wozniak and others sign an open letter calling for a six-month pause in training AI more powerful than GPT-4.
- 2023.03.15Midjourney releases v5; days later an AI-generated image of the pope in a puffer jacket spreads as if it were real.
GPT-4Introducing ClaudePause Giant AI Experiments: An Open Letter
Intelligenza artificiale: il Garante blocca ChatGPTChatGPT è di nuovo raggiungibile dall’Italia
Pause Giant AI Experiments: An Open Letter
New Midjourney V5 Generative AI Offers Photorealistic Images and Hands With 5 FingersAI-generated images of Pope Francis in puffer jacket fool the internet
2023.02.07 · Big Tech Strikes Back
Bing, Bard, LLaMA
Google, out to defend search, and Microsoft, out to take it, launched chatbots in the same week. Meta released its model to researchers.
- 2023.02.24Meta releases LLaMA, a model for research; its weights leak within a week, setting off a boom in open models that run on personal computers.
- 2023.02.07Microsoft unveils a new Bing built on GPT, and the emotional conversations of its code name “Sydney” draw attention.
- 2023.02.06Google announces its chatbot Bard; two days later a wrong answer in a promotional demo comes to light and Alphabet loses about $100 billion in market value.
An important next step on our AI journeyReinventing search with a new AI-powered Microsoft Bing and Edge, your copilot for the webIntroducing LLaMA: A foundational, 65-billion-parameter language model
Introducing LLaMA: A foundational, 65-billion-parameter language modelRunaway LLaMA: How Meta’s LLaMA NLP model leaked
Reinventing search with a new AI-powered Microsoft Bing and Edge, your copilot for the web
An important next step on our AI journeyGoogle shares tank 8% as AI chatbot Bard flubs answer in ad
2023.01.13 · Copyright
Who Owns an AI Image?
In the summer of 2022, an AI-made picture took first place in an art competition at a U.S. state fair, and an image generator that anyone could download and run was released. The next year, artists and a photo agency sued, and the U.S. Copyright Office and a photography contest declined to treat AI-made images as human work.
- 2023.04.13German artist Boris Eldagsen refuses the Sony World Photography Awards prize he won in the creative category with an AI-made image, saying “AI is not photography,” while the organizers say he misled them about how much AI was involved.
- 2023.02.21The U.S. Copyright Office finds that the Midjourney-generated images in the comic Zarya of the Dawn are not the product of human authorship, while its text and arrangement stay protected.
- 2023.02.03After filing in London in January, Getty Images also sues Stability AI in federal court in Delaware, alleging that it copied more than 12 million photos along with their captions.
- 2023.01.13Three artists file a class action against Stability AI, Midjourney and DeviantArt, claiming their work was used for training without permission.
Artificial intelligence artwork wins 1st place at Colorado State Fair, causing controversyAI Art Generators Hit With Copyright Suit Over Artists’ ImagesGetty Images StatementGetty Images Lawsuit Says Stability AI Misused Photos to Train AIZarya of the Dawn (Registration # VAu001480196)
Photographer refuses prestigious award after admitting image was AI-generatedGerman artist refuses award after his AI image wins prestigious photography prize
Zarya of the Dawn (Registration # VAu001480196)
Getty Images StatementGetty Images Lawsuit Says Stability AI Misused Photos to Train AI
AI Art Generators Hit With Copyright Suit Over Artists’ Images
2022.11.30 · Starting Point
The Launch of ChatGPT
OpenAI released a conversational chatbot built on GPT-3.5 as a free research preview. Typing a question into a box became what it meant to use AI.
- 2023.01.23Microsoft announces a further multiyear, multibillion-dollar investment in OpenAI.
- 2023.01UBS estimates 100 million monthly users two months in, the fastest-growing consumer app at the time.
- 2022.12.21Google executives are reported to have declared a “code red” over the threat to the search business.
- 2022.12.05ChatGPT passes one million users five days after launch.
Introducing ChatGPTChatGPT reaches 100 million users two months after launchMicrosoft and OpenAI extend partnership
Microsoft and OpenAI extend partnership
ChatGPT reaches 100 million users two months after launch
Report says Google is panicking over threat posed by OpenAI’s chatbot
ChatGPT gained 1 million users in under a week. Here’s why the AI chatbot is primed to disrupt search as we know it
2026 The Year of Delegation and Control
As people handed work to AI, stock markets and governments reacted, and major labs prepared to go public while disclosing research acceleration risks and compute costs. Cyber capabilities and agent incidents made pacing and isolation security priorities. A researcher who left Anthropic warned of the risks before a city council and on television. Mathematicians debated standards for evaluating AI results, and specialized life sciences models arrived. Google opened a world model to subscribers, and Meta held back the weights of its first Muse model. The U.S. administration began calling AI “Super Intelligence,” and local voters backed limits on data centers. Website training controls, protections for children and conversations on companion chatbots, and EU AI Act enforcement followed. Chinese AI firms went public, the U.S. and China clashed over distillation, and compute infrastructure expanded through optical connectivity and tests in orbit.
2025 The Year of Agents and Compute
It opened with the DeepSeek shock, and coding agents changed how software gets built. Data center deals grew to the gigawatt scale, and Korea was promised 260,000 GPUs. Chinese open models contended for the lead, and chips became leverage between the U.S., China and the Gulf. Courts ruled for the first time on whether AI training is fair use, and U.S. states and Washington clashed over who writes the rules.
2024 The Year of Senses and Reasoning
AI began to make video, hold spoken conversations and think before answering. Two Nobel Prizes went to AI research, and at the end of the year AI took direct control of the computer screen. A year of elections began with a fake voice, a price war broke out among China’s models, and Big Tech turned to nuclear plants to power its data centers.
2023 The Year the Race Began
Having seen ChatGPT, Big Tech rushed out chatbots of its own. As model weights were released, anyone could run a language model, and the people who built AI warned of its risks. Copyright suits over training data began, China opened only approved chatbots to the public, and the U.S. tightened its controls on AI chip exports.
