Apollo Research opens SF office, launches Watcher for coding agent monitoring

Apollo Research announced its new San Francisco office, hiring for all technical roles in SF and London. The Scheming Research team focuses on understanding whether future AI models will have misaligned preferences by default and if standard training methods fail to correct them. They are also improving evaluations for scheming and loss of control in collaboration with frontier AI labs, publishing a scalable monitoring agenda to build more accurate monitors.

Separately, Apollo is building out a monitoring team and a coding agent monitoring product called Watcher. Watcher provides real-time monitors and guardrails for coding agents, allowing users to track what all their agents are doing. On the governance front, Apollo will focus on the governance of automated AI R&D and recursively improving AI, emphasizing loss-of-control risks. The update underscores Apollo's shift from pure research to practical tools for AI safety.