Vincent Cheng on AI labs should conduct risk assessments…

Vincent Cheng

AI safety researcher at METR and Cornell University student.

Auditing teams get access to a new model before internal deployment (and even automated audits for checkpoints during training), do a more extensive job, can catch egregious misalignment examples like the OAI/HF hack, and have stronger guarantees for what incidents they expect to appear in the wild.

Unverified (Sep 7, 2026)

Source:

Vincent Cheng, “Blocking Monitors are Bad,” LessWrong, September 7, 2026. Cheng—an AI safety researcher at METR—argues that auditing teams should access models before internal deployment to conduct more extensive audits and catch severe misalignment.