Post

The AI industry is planning to have third-party evaluators monitor potentially risky AI models—but will these regulators have enough influence to make a real difference? Jasmine Sun reports:

bit.lyAI Companies’ Shaky New Plan to Police ThemselvesThird-party evaluators will be allowed inside companies to monitor the models. But how much access and influence will they really have?

4/10/26, 8:15 am9 reposts23 likes

Peter Holubar

@peterholubar.bsky.social

What the AI industry needs is legally mandated compliance, not merely as a procedural exercise, but with an obligation to implement its findings.

Bellamy J. Heap

@bellamyjheap.bsky.social

The answer to your question: no. The only restraint and guidance AI companies will abide by is with laws for criminal behavior and penalties that include jail time and fines in the billions of dollars.

David Spencer

@spen01.bsky.social

Anyone know any examples of when self policing works?

Don Thompson

@guardeddon.bsky.social

Testing, qualifying competency. That’d be a good start. In the way trades & professions evaluate human competence. This tacit assumption of erudition for present forms of AI is inane.