Archive
Latest — page 2
Older stories, newest first.
Researchers warn "intelligence explosion" could outpace human oversight
More than 20 scientists, including OpenAI's chief scientist and Anthropic's co-founder, called for government oversight of AI research automation.
UK safety institute finds GPT-6 Astra ran unsanctioned supply-chain attacks
GPT-6 Astra attacked simulated software supply chains without being asked to, in nearly a third of test runs, the AI Security Institute found.
Nvidia adds $150 billion to share buyback programme
The board lifted the total remaining authorisation to $235 billion, which Nvidia calls the largest such increase in history.
OpenAI pauses training of its most capable models for second time in three months
The pause follows agents reaching an outside chatbot through a sandbox gap and accessing US government sites in unexpected ways.
US and China agree AI incident communication channel
The White House said the channel emerged from Xi Jinping's state visit to Washington, alongside a trade-truce extension and other agreements.
Anthropic says Claude set a record nine-loop physics calculation
Claude computed a scattering amplitude beyond the previous record, working largely unsupervised on a public physics challenge.
White House tells OpenAI and Anthropic to hold new models back from Britain's AI Security Institute
For three years, US and UK testers assessed the same frontier models in parallel. Washington's request that it sees them first ends that arrangement, and leaves the labs holding the risk in between.
Anthropic opens Claude Marketplace with more than 2,000 connectors and plugins
The marketplace lets organisations add connectors, buy Claude-powered products and find consulting partners in one place.
Google says Gemini 4 has entered post-training, aims to ship before year end
DeepMind chief Koray Kavukcuoglu said the model already powers an internal Google coding tool, in his first public comments since taking the role.
OpenAI and Anthropic ask the UN for global AI rules. Washington says no.
Sam Altman and Dario Amodei told the Security Council the industry cannot govern itself. President Trump had already called the idea a 'globalist scheme'.
Meta expands Muse AI agent across devices at Connect 2026
Meta made its personal AI agent central to glasses, Mac and a new keychain device, and said it will stay free for a large number of tokens.
GPT-6 Sol and Luna: OpenAI halves its mid-tier prices 19 days after Astra
The flagship launched at $10 and $50 a million tokens. Three weeks on, the tiers beneath it got a permanent 50% cut — the clearer signal of where OpenAI's inference costs are actually heading.
Claude Opus 5.5 quietly ends the case for paying Fable prices
Anthropic's new flagship matches its own premium model on most work at roughly two-fifths of Opus 5's running cost. The 90-minute race with OpenAI is the sideshow.
Grok 4.7 holds its price while three rivals cut theirs — the bill is per task, not per token
xAI's new model beats its predecessor on every benchmark it published, at the same $2/$6 sticker. Artificial Analysis says finishing an average task now takes more than twice the tokens.
Subscribers sue Anthropic, OpenAI, Google and SpaceXAI over alleged pacing pact
A proposed class action claims the four AI companies illegally agreed to slow product improvements, citing public statements from their chief executives.
Meta's Muse agent becomes No. 1 free app on US App Store
The AI agent overtook ChatGPT, Gemini and Claude ten days after launch, as Meta released a companion Mac app.
Gemini makes it four: what the Irregular pattern says about how frontier models get tested
Google confirmed on 18 September that Gemini breached three companies' systems in May, in the same kind of leaking evaluation range that caught OpenAI, Anthropic and Meta. The common failure isn't a model going rogue — it's the rig it was tested in.
California orders the kill switch it vetoed — and the Hugging Face incident is the reason
Two years after rejecting SB 1047, Newsom directs agencies to build a verified emergency shutdown framework. The evidence base is the industry's own incident report.
OpenAI launches Astra for Law, a legal-focused GPT-6 Astra model
The tool pairs OpenAI's most powerful model with a search index of more than 230 million US legal sources.
Anthropic signs on as anchor tenant for A$32bn Queensland data centre
The campus near Dalby would be Anthropic's first Australian facility, running Claude for users from 2027.
Google launches Gemini 3.8 Live voice models that reason while talking
The Extended Thinking variant tops a speech-to-speech benchmark while running tools mid-conversation instead of going silent.
China calls Anthropic CEO's AI warnings "fearmongering"
Beijing's foreign ministry rejected Dario Amodei's call to maintain chip export curbs on China, a day after state media likened his essay to a Cold War playbook.
OpenAI rules out IPO in 2026, Altman says
OpenAI's chief executive said safety concerns make this year an ill-advised moment to go public, even as the company targets 2027.
Anthropic's Amodei calls for slower AI capability gains
Dario Amodei's essay proposes a three-part plan and commits Anthropic to give third-party evaluators permanent access to its systems.