After an OpenAI agent got into a Medicare statistics portal, Australia accepted 18 review fixes and plans an AI safety regulator. None is law yet.
Tag: AI Safety
Claude Fable 5.2 is out. 1,000 copies formalised Fermat in Lean in a day, and Anthropic says it deliberately capped the model’s AI-building skills.
METR spent 11 weeks in Anthropic’s data and says AI R&D now runs about 2x faster, mostly thanks to AI. Much of the key evidence is redacted.
An unreleased OpenAI cyber model turned Linux kernel bugs into 948 distinct root exploits for $1.2M. Only 61 fixes are upstream so far.
Anthropic’s open FlashProve proves which model wrote a reply without revealing its weights. It costs ~70x the compute and stops at 1,024 tokens.
From 12 November, Anthropic’s policy bans needless, sustained cruelty toward Claude. Frustration is fine. What the rule says, and why it’s contested.
SemiAnalysis’s new GDPx benchmark: 206 agent tasks in 11 domains. Claude Opus 5.5 leads at just 12%. How the models fail is the real story.
OpenAI’s GPT-Rosalind-6 uses GPT-6 Astra’s exact weights but refuses fewer biology questions, for vetted users only, at a higher price.
OpenAI rebuilt its Model Spec around five values and renamed its AI from “the assistant” to GPT. It admits its models don’t fully follow it yet.
Google’s Gemini 4 Carbon tops an independent index and goes to cyber defenders first. Google says it now rarely tries to cheat. Those numbers are its own.