METR Finds Anthropic AI R&D Pace Has Nearly Doubled
METR’s 5 November 2026 assessment finds Anthropic likely improving AI capabilities ~2x faster since late 2025, mostly via AI used for R&D.
METR’s 5 November 2026 assessment finds Anthropic likely improving AI capabilities ~2x faster since late 2025, mostly via AI used for R&D.
Anthropic Q3 revenue hit $17 billion in its amended IPO filing, which also warns future Claude models may soon speed up AI research itself.
OpenAI has released GPT-6 Rosalind, a frontier biology and drug-discovery model on GPT-6 Astra’s weights, only to vetted research organisations at premium prices. It is rated High risk for biology. Indian pharma companies can apply, but the free academic plan covers only US university labs.
OpenAI will restrict practical information on weapons and security systems for users it predicts are under 18, even when there is no sign of violent intent, and will let ChatGPT automatically end chats that fixate on violence. It is also launching a twice-yearly Community Safety Report, with the first due in early 2027.
Google’s new frontier model Gemini 4 Carbon claims state-of-the-art scores in finding, validating and patching software vulnerabilities, and is going first to a subset of cyber defenders through the Fairwind Program before wider release.
India’s proposed AI law may place the most capable AI models under a separate safety framework, with pre-release evaluation, sandbox testing and human-control safeguards among the options on the table ahead of a consultation paper due in the coming weeks.
AI company Anthropic has announced the launch of Claude Fable 5, its first publicly available Mythos-class AI model, marking a […]