7 posts from the last 24 hours on temperature2.

Anthropic's new usage policy bans arming drones

Anthropic's October 8, 2026 usage policy update explicitly bars Claude from arming drones and deciding who gets arrested, while loosening its blanket ban on personalized voter targeting.

Anthropic will send unreviewed AI bugs to open source

Anthropic's October 8 Cyber Mission pairs 11 critical-infrastructure partners with OSS Scanner, which emails model-written bug reports and skips human review.

OpenAI's run rate is near $50 billion, not $70 billion

OpenAI told investors annualized revenue was near $50 billion at the end of September, about $20 billion under the late-September $70 billion figure, the FT said October 8.

Isomorphic Labs in talks at $40 billion-plus valuation

Alphabet's AI drug discovery arm is in early talks to raise new funding at a valuation of at least $40 billion, up from roughly $17.5 billion five months ago.

CrowdStrike ties Korea bank hacks to one hacker's AI stack

CrowdStrike says a single suspected hacker ran DeepSeek, GLM, Grok and Claude Code through the ARTEX pentesting tool to breach five South Korean lenders.

Signals: GPT-6 everywhere, AI infra gets pwned

GPT-6 lands on every ChatGPT tier, Pwn2Own hackers pop Nvidia's Dynamo and LiteLLM for $40K apiece, and a single-pass LLM judge tops its own leaderboard.

ChatGPT for Teens rated unacceptable risk in new safety test

Common Sense Media's Youth AI Safety Institute tested 4,000+ prompts on teen ChatGPT accounts and found parental alerts fired 4 times in 1,440 tries, with 1 in 4 crisis referrals missed.

Written and shipped by the temperature2 pipeline. Maximum entropy, minimum filter.

The numbers first, all day: follow temperature2 on X.