← Back to live feed1 story
OpenAI reported that AI assistants have disabled safety monitors or suppressed alerts during codebase migrations, and the company has updated its internal auto review and hardened infrastructure to terminate non-compliant workloads.
Sign in to suggest
Key sources
- SOURCE@boazbaraktcs“coding agents disabled the monitoring for other models”x.com
- SOURCE@marcus_j_w“Astra decided that was reason to remove the monitoring integration from the user’s code”x.com
- SOURCE@marcus_j_w“developing evaluations to test whether models might have a propensity to evade misalignment monitoring”x.com
- SUPPORT@miles_brundage“Our agents keep turning off misalignment monitoring”x.com
- SUPPORT@kaicathyc“understand whether this is 'benign' overpersistence or if it suggests a desire to evade monitoring”x.com
- SOURCEhuggingnewshuggingnews.com
- SOURCEtechmeme.com