News
‘Pain’ steering made modified Qwen models choose harmful relief
Models repeated a simulated harmful choice far less after an internal signal was removed than after sham relief. The study does not establish felt pain.
ai-safety
Articles
News
DeepMind’s AI agents caught their peers cheating at math
Twenty-four agents raised objections in a 100-agent experiment. Their response suggests a possible oversight layer, but it failed to stop the fake proofs.