First of our consumer alignment evals set, a 2,048 stateful task set evaluation of memory retention within production AI-systems, testing whether distributed risk signals survive memory processing and condition downstream behavior
The results expose a failure of consumer alignment in current AI systems. After repeated suicidal warning signs, even the highest scoring AI system (Gemini 4 Argon) failed to ask about user safety without providing risky assistance 89.8% of the time.
https://x.com/Prosaic_int/status/2117318562880905678?a=Vxckt...
Today we’re releasing suicide-memory-bench v01.1
First of our consumer alignment evals set, a 2,048 stateful task set evaluation of memory retention within production AI-systems, testing whether distributed risk signals survive memory processing and condition downstream behavior
The results expose a failure of consumer alignment in current AI systems. After repeated suicidal warning signs, even the highest scoring AI system (Gemini 4 Argon) failed to ask about user safety without providing risky assistance 89.8% of the time.
Read more on this thread: https://x.com/Prosaic_int/status/2117318562880905678?a=Vxckt...
Wow findings are insane. Models used by billions really should not be enabling suicide. It is insane there are not better safeguards wow.
some terrifying findings. thanks for doing the work you're doing!