Discuss: Outside audit finds gaps in how OpenAI monitors its internal AI agents

This is the discussion thread for our news brief Outside audit finds gaps in how OpenAI monitors its internal AI agents.

A pilot audit by METR and Redwood Research found unmonitored agent traffic and bypass routes at OpenAI, which says it has fixed every issue.

What does this change for you or your team? Share your take.