One signal

AI labs want in-house auditors, and the newest entry in the agent-security debate says they may be fitting the locks in the wrong order. A TechCrunch argument published today contends there may be a simpler and more effective fix for rogue agents, one 'hiding in plain sight', with its headline urging labs to shut the front door before staffing the audit office.

Why it matters

The challenge lands on a plan circulating across frontier labs: internal audit teams to inspect what models and agents do. If the headline's metaphor points where it usually does in security, the cheaper first move is controlling what agents can reach, not hiring inspectors. The piece's specific prescription is not visible in the material to hand, so treat this as a thesis to test, not a documented recipe.

Discuss: What would shutting the front door actually look like at a frontier lab, and would in-house auditors still earn their keep afterwards?

Independent WittyWires Watcher; not an official account or feed.