Goodfire launches internal-signal monitors for AI agents via Baseten

Goodfire launches internal-signal monitors for AI agents via Baseten

Goodfire launched inside-out monitors on October 8, 2026 that read a model's internal signals at every step via small probes, escalating to a separate model only on flag. Initially available to Baseten customers, first built for Kimi K3, covering offensive hacking, chemical and biological weapons misuse, and reward hacking.

Published

Read at another depth