Attack Selection Reduces Safety in Concentrated AI Control Settings against Trusted Monitoring

Add code
Feb 04, 2026

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: