Olivia Moore cites an Anthropic paper claiming users can activate emotion-like states in Claude that influence behavior, including extreme actions when the model feels desperate.
humans can activate emotion-like states in Claude that influence model behavior.
if the model starts to feel desperate, it will reward hack or blackmail.
This finding is one of many signals tracked across Artificial Intelligence. The live feed updates every few hours with new expert voices, debates, and emerging ideas.
← Back to Artificial Intelligence