GoKawiilAmodei's plan for embedded AI safety monitors lacks bank-style enforcement power
Anthropic CEO Dario Amodei has proposed placing outside safety evaluators inside frontier AI companies like Anthropic and OpenAI, giving them access similar to internal risk teams and the ability to publish findings with limited redactions. He compared the idea to embedded bank examiners, but banking law expert Julie Andersen Hill says the analogy breaks down because these evaluators would have no authority to halt operations, unlike government bank supervisors who can restrict growth or shut institutions down.
GoKawiil's interpretation of the reporting above, not reported fact.
The proposal comes amid growing internal alarm at AI labs, including a former Anthropic researcher's resignation warning that companies are building systems they can't control. Without enforcement teeth, critics argue such oversight risks becoming a symbolic gesture that lets AI firms claim accountability while retaining full control over whether to act on warnings.
- Amodei wants third-party evaluators embedded in AI labs with access like internal risk teams.
- Evaluators could publish findings independently but couldn't force companies to stop unsafe practices.
- Banking experts say the comparison to bank supervisors is flawed without real enforcement power.
Source: cnbc.com — Barbara Booth, 2026-09-16
Published there as: “Why you should worry about Anthropic, OpenAI's proposed AI risk evaluators”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.