Discussion and research on AI alignment. alignmentforum.org
Google DeepMind and UK AISI present a safety case sketch for control monitoring of frontier agent deployments, comparing asynchronous, semi-synchronous and fully-synchronous monitors on oversight, latency and harm-prevention conditions.
Argues AI safety research must plan for deploying known scheming models, analysing how catching escape attempts lets developers reduce risk via classifiers, policy training, and monitoring, and recommends a responsible disclosure and third-party deployment policy.
Sketches a structured safety case arguing an internally deployed LLM agent won't exfiltrate data, based on a red-team control evaluation of monitoring protocols and extrapolation to deployment.