Goodfire says its new ‘inside-out’ monitors catch rogue AI agents at a fraction of the cost

Goodfire launched 'inside-out' monitors that inspect AI agents' internal states in real time, reducing the need for expensive second-AI oversight. The system only triggers external checks when suspicious activity is detected, cutting costs significantly. This approach offers a more efficient safeguard for AI agent deployment, addressing safety concerns without the overhead of continuous external monitoring.