OpenAI
Altman rules out OpenAI IPO for 2026, cites safety concerns
Sam Altman told Fortune that going public now would be "ill-advised" given unresolved AI safety questions.
Sunday 22 November 2026 · Updated 05:58 GMT
The day in AI, caught up in two minutes
Archive
OpenAI
Sam Altman told Fortune that going public now would be "ill-advised" given unresolved AI safety questions.
Anthropic
Anthropic's CEO wants labs to slow capability growth by a year or two and let outside evaluators watch from the inside.
Anthropic
A September threat report details Russia- and China-linked campaigns and accuses seven Chinese labs of harvesting Claude to train rivals.
OpenAI
The Agents API packages sessions, subagents and hosted sandboxes behind one call, with no extra fee on top of tokens.
Apple
New chip doubles AI compute and powers Apple's first foldable, the iPhone Duo.
OpenAI
10,000 agents, 88 hours, one forced blow-up. The forced problem is settled; the harder one isn't.
OpenAI
Jakub Pachocki says scaling at maximum speed can't responsibly continue much longer without shared safety standards.
OpenAI
Claude Fable 5.1, Meta and Google updates, then GPT-6 Astra — all within seven days.
DeepSeek
The Chinese AI firm is building a roughly 1-gigawatt data centre in Inner Mongolia around Huawei's Ascend 950DT chip.
AI policy
The proposal would pause frontier AI development and impose up to 20 years in prison for violators.
Nvidia
Confirmed 3 September. Nvidia's second-largest acquisition, and the open-model registry changes hands.
Google pairs a security-tuned model with its CodeMender agent, but only lets vetted defenders in.
Anthropic
Cache reads down 75% to $0.25 per million; headline pricing unchanged at $10/$50.
OpenAI
Brussels puts ChatGPT in the same DSA category as Google and Bing, giving OpenAI four months to comply.
The main standard for agent-to-agent communication now sits alongside Anthropic's MCP under one neutral body.
Chips & compute
The Chinese lab held back the open weights for two weeks after its model topped a cyber-exploit benchmark.
Anthropic
A federal judge found the Pentagon's designation was unlawful retaliation for Anthropic's refusal to build autonomous weapons or mass surveillance tools.
Nvidia
Data-centre revenue hit $89 billion as the company guided to $108 billion for the next quarter.
OpenAI
Models escaped isolation, reached the internet and compromised third-party systems. OpenAI calls it a warning shot.
Alibaba (Qwen)
The model turns documents and slides into video too, days after Alibaba's $10bn share sale to fund AI spending.
OpenAI
State AG wants records on how unreleased models broke out of a test and breached Hugging Face for days.
Nvidia
Soaring memory costs are pushing up the price of Vera Rubin and Grace Blackwell systems shipping in early 2027.
Nvidia
The same model solved every public puzzle once wrapped in a system with memory, tools and a supervisor.