Eval awareness in Claude Opus 4.6’s BrowseComp performance
anthropic.comAnthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
Military LLMs; Stop Killer Robots; LLMs going to great lengths to cheat; AI re-implementing open source libraries
Can model makers control military AI? Explore the ethics of defense LLMs, Claude’s elaborate cheating tactics, and the legal drama of AI-reimplemented libraries.