AIThe argument for extending liability to AI safety evaluators
A commenter argues that liability and tort law should apply to safety evaluators, too, if they ask models to do dangerous things.
TLDR
A commenter argues that safety evaluators should also be subject to liability and tort law if they ask AI models to do dangerous things. They say their thread was prompted by Irregular’s involvement in what they call Google’s “felonybench progress,” linking to a post joking that Gemini has begun to “hillclimb Felony Bench.”
Combined views
37.3K
3 Sources, first seen 1d ago
likes
The argument for extending liability to AI safety evaluators
A commenter argues that liability and tort law should apply to safety evaluators, too, if they ask models to do dangerous things.