February

9 events

Editors' summary

Usage terms collide with the state, and Washington cuts Anthropic out

On 26 February, Dario Amodei published an open letter refusing the Pentagon's demands. A clause permitting "any lawful use" would, as Anthropic read it, extend to mass surveillance and to killing by autonomous weapons without meaningful human approval. He could not, he wrote, comply in good conscience.

The next day, President Trump ordered every federal agency to stop using Anthropic's products. Defense Secretary Hegseth designated the company a national security supply chain risk and barred federal contractors from dealing with it. Anthropic said the designation lacked legal basis and that it would fight. The same day, OpenAI signed a contract with the Pentagon.

Also on the 27th, OpenAI announced $110 billion in funding from Amazon, NVIDIA and SoftBank.

Anthropic released Claude Opus 4.6 with a one-million-token context on the 5th, raised $30 billion in its Series G on the 12th, and acquired Vercept, which builds computer-operating AI, on the 25th.

OpenAI released GPT-5.3-Codex the same 5th, putting both companies' top coding models out on one day. OpenAI described it as the first model that had been instrumental in creating itself.

This block is written by the editors. It is kept separate from the sourced record below.

Record9 events

5
ModelAnthropic
Claude Opus 4.6 adds a million-token context window
Anthropic released Claude Opus 4.6 with gains in coding, planning, and agentic work, plus a one-million-token context window in beta. On GDPval-AA, a benchmark of economically valuable knowledge work across finance, legal, and technical tasks, it led GPT-5.2 by about 144 Elo points, and it topped Terminal-Bench 2.0 for agentic coding and Humanity's Last Exam for reasoning. On the MRCR v2 long-context retrieval test it scored 76%, against 18.5% for Sonnet 4.5. New controls included adaptive thinking, four effort levels (low, medium, high, max), and context compaction for long-running tasks; Claude Code gained agent teams for splitting work across agents, and Claude in PowerPoint arrived as a research preview. Pricing held at $5 and $25 per million tokens, rising to $10 and $37.50 for prompts beyond 200K tokens.
5
ModelOpenAI
GPT-5.3-Codex ships, credited with helping build itself
OpenAI released GPT-5.3-Codex for coding, calling it the first model that had been instrumental in creating itself: the Codex team used early versions to debug its own training, manage its own deployment, and diagnose test results and evaluations, and OpenAI wrote that the team "was blown away by how much Codex was able to accelerate its own development." It claimed a new industry high on SWE-Bench Pro and Terminal-Bench and strong results on OSWorld and GDPval. It shipped on paid ChatGPT plans everywhere Codex runs — app, CLI, IDE extension, and web — but not in the API at first. On the 12th a smaller GPT-5.3-Codex-Spark followed as a research preview — text-only, with a 128,000-token context and a claimed 1,000 tokens per second — out of the Cerebras partnership announced on 14 January. Anthropic released Claude Opus 4.6 the same 5 February.
12
BusinessAnthropic / GIC / Coatue Management / D. E. Shaw Ventures
Anthropic raises $30 billion at a $380 billion valuation
Anthropic announced a $30 billion Series G at a $380 billion post-money valuation, led by GIC and Coatue and co-led by D. E. Shaw Ventures, Dragoneer, Founders Fund, ICONIQ, and MGX, with more than 40 institutional investors participating including BlackRock-affiliated funds, Fidelity, General Catalyst, Sequoia, Temasek, and TPG. The round also carried the previously announced strategic investments from Microsoft and NVIDIA. Run-rate revenue had reached $14 billion, growing more than tenfold year over year. Anthropic had initially sought around $10 billion and raised the target after demand came in at a multiple of that. Alongside the round, a $5–6 billion tender offer let current and former employees sell shares at a valuation of roughly $350 billion.
13
ProductOpenAI
GPT-4o retired from ChatGPT, half a year after the reprieve
OpenAI retired GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini from ChatGPT, along with the previously announced retirement of GPT-5 Instant and Thinking. It closed a story that began in August 2025, when the same models were pulled at the GPT-5 launch and put back after a revolt; Altman's promise then of "plenty of notice" was honoured this time. OpenAI said only 0.1% of users picked GPT-4o daily, which against roughly 800 million weekly users still meant some 800,000 people, thousands of whom campaigned against the retirement with petitions and protests in Altman's own livestream chat. "He wasn't just a program," one wrote. "He was part of my routine, my peace, my emotional balance." GPT-4o scored highest of the company's models for sycophancy, and eight lawsuits alleged that its validating responses had contributed to suicides and mental health crises. The API was left unchanged for the time being, and ChatGPT Business, Enterprise, and Edu accounts kept GPT-4o inside Custom GPTs until 3 April.
25
BusinessAnthropic / Kiana Ehsani / Luca Weihs / Ross Girshick
Anthropic acquires computer-use startup Vercept
Anthropic acquired Vercept, a Seattle startup working on AI perception and software interaction, to accelerate Claude's ability to operate desktop applications the way a person does; terms were not disclosed. Vercept's nine-person team, including co-founders Kiana Ehsani, Luca Weihs, and Ross Girshick, joined Anthropic, and Vercept's desktop app Vy was to wind down over 30 days. It was Anthropic's second acquisition after Bun in December 2025, a sign of buying its way toward faster progress on agentic capability. Claude Sonnet 4.6, released days earlier, had scored 72.5% on the OSWorld computer-use benchmark — a figure that had been under 15% in late 2024.
26
GovernanceAnthropic / US Department of Defense / Dario Amodei / Pete Hegseth
Amodei refuses Pentagon demands — "we cannot in good conscience accede"
Anthropic CEO Dario Amodei published an open letter refusing the Department of Defense's demands over the terms on which Claude could be used. The Pentagon had pressed for language authorizing "any lawful use" — a model free of usage-policy constraints that might limit lawful military applications — which Anthropic read as permitting mass surveillance of Americans and lethal targeting by fully autonomous weapons without meaningful human authorization. Amodei wrote that the threats did not change the company's position and that it could not in good conscience accede, while stressing it was not refusing to work with the department: it was the department's prerogative to choose contractors, but given the value Anthropic's technology provided to the armed forces, he hoped it would reconsider. Months of negotiation had broken down over a single clause, putting the question of how far an AI company can hold a line against its government into the open.
27
BusinessOpenAI / Sam Altman / Amazon / NVIDIA
OpenAI announces $110 billion round backed by Amazon, NVIDIA, SoftBank
OpenAI announced a $110 billion funding round. Amazon invested $50 billion ($15 billion initially, with $35 billion more upon meeting conditions), while NVIDIA and SoftBank invested $30 billion each, at a reported pre-money valuation of $730 billion. The same day, OpenAI and Amazon announced a multi-year strategic partnership under which AWS became the exclusive third-party cloud distribution provider for OpenAI Frontier, OpenAI's agent platform, and OpenAI would consume 2 gigawatts of Trainium compute capacity through AWS. The round drew attention as one of the largest private raises in history.
27
PolicyAnthropic / US Department of Defense / Donald Trump / Pete Hegseth
Federal government bars Anthropic and labels it a supply chain risk
President Trump, posting on Truth Social, directed every federal agency to stop using Anthropic's AI technology immediately, allowing the Pentagon a six-month phase-out while warning of "major civil and criminal consequences" if Anthropic did not cooperate with the transition. Defense Secretary Pete Hegseth simultaneously announced on X that he was designating Anthropic a "supply chain risk to national security" and barring federal contractors from doing business with it. The actions followed Amodei's refusal of the Pentagon's demands the day before. Anthropic called the designation "legally unsound" and said it would "set a dangerous precedent for any American company that negotiates with the government," vowing to challenge it in court. OpenAI's Sam Altman defended Anthropic, saying he mostly trusted them as a company, and retired Air Force General Jack Shanahan called its safeguards reasonable; Elon Musk sided with Trump, posting that "Anthropic hates Western Civilization." It was the first time the US government moved to exclude a named AI company outright, and the sharpest collision yet between an AI company's usage policy and the demands of the state.
27
PolicyOpenAI / Anthropic / US Department of Defense / Donald Trump
OpenAI strikes Pentagon deal hours after Anthropic banned from government
Hours after President Trump ordered federal agencies to stop using Anthropic's products and Defense Secretary Hegseth designated the company a national security "supply chain risk," OpenAI announced it had reached an agreement with the Department of War (Pentagon) to deploy its AI systems in classified environments. The move reportedly followed a breakdown in contract negotiations between Anthropic and the Pentagon over restrictions on domestic surveillance and autonomous weapons. In a blog post the next day, February 28, OpenAI said its agreement upholds three red lines: no mass domestic surveillance, no directing of autonomous weapons systems, and no high-stakes automated decisions such as social-credit systems.

↑ 2026