Anthropic accused of backtracking on AI safety commitments
An article on LessWrong alleges the company released a model it classified as risky without meeting requirements it had previously promised.
TLDR
An article on LessWrong accuses Anthropic of backtracking on safety commitments, alleging it released a model it classified as risky without meeting previously promised requirements. A commenter also claims Anthropic once committed not to advance the AI capability frontier, then rationalized triggering an arms race it believed posed an existential risk.
Combined views
444
1 Source, first seen 15d ago