Goodfire launches low-cost 'inside-out' monitors to catch rogue AI agents

Goodfire, a startup focused on AI interpretability, has launched a new monitoring system for AI agents that reads signals from inside a model while it works, rather than having a separate AI review its outputs.

The monitors are now available to customers of Baseten, a model hosting and inference platform. They can be configured to watch for risks including offensive hacking, chemical and biological weapons misuse, and reward hacking, with automated responses such as logging the event, sending it for human review, or refusing the request.

The system works like airport security: small detectors called probes read the model's internal signals at every step of an agent's work, and only when a probe flags something does a separate AI model take a closer look. Because the probes reuse computations the model is already performing in its forward pass, Goodfire says the approach is much cheaper to run than traditional monitors that reread everything the monitored model outputs.

In tests on Kimi K3, monitoring around 1,500 sessions cost roughly $51, compared with $233 for a cheaper AI model checking every step and about $10,000 for a top-tier model. The probes caught 94% of malicious hacking sessions and flagged 8.7% of harmless ones for a second look. Running four probes at once added less than 2% to the time before the model starts responding.