Microsoft
Scheming risk prevention 2%
0%
Safety case: process
0%
Safety case: accountability
0%
Internal deployment protocol
0%
External deployment protocol
0%
Plan for if an AI is caught scheming
0%
Training: remove scheming capabilities
0%
Training: adversarial training
50%
Training: safe architecture
AIs show signs that if they were more capable, they would sometimes scheme, i.e. fake alignment and subvert safety measures in order to gain power. AI companies should prepare for risks from models scheming, especially during internal deployment: if they can't reliably prevent scheming, they should prepare to catch some schemers and deploy potentially scheming models safely.
Safety case: process
0%Click to show details/rubric
Safety case: accountability
0%Click to show details/rubric
Internal deployment protocol
0%Click to show details/rubric
External deployment protocol
0%Click to show details/rubric
Plan for if an AI is caught scheming
0%Click to show details/rubric
Training: remove scheming capabilities
0%Click to show details/rubric
Training: adversarial training
0%Click to show details/rubric
Training: safe architecture
50%Click to show details/rubric