What did the UK's Joint Committee on Human Rights propose for AI systems?

The UK's Joint Committee on Human Rights proposed a statutory body with power to test an AI system and withdraw it from the market before release. The UK government must respond to the committee's recommendation by 14 November 2026, and the proposal would also shift liability toward model builders if a system fails.

What did New York State do to address frontier AI safety?

New York opened a $55 million center staffed with more than 80 scientists to write standards and test frontier AI models. It is one of three pre-market safety efforts this week, alongside the UK's proposed withdrawal power and France's Paris Call for AI Safety and Security, all aimed at models before they ship.

Why does Yves Mulkers say pre-market AI regulation does not stop a live agent from failing?

Yves Mulkers argues pre-market gates like the UK, New York and France efforts act on the model before deployment, not on an agent already running inside a company with credentials and data. He cites Microsoft's reported Copilot discount, 30 to 50 percent off for a 1,000 to 10,000 seat commitment, as pricing how hard a system is to stop.

UK Human Rights Committee Can Withdraw AI

The Argument · Ins7ghts Daily, Thursday 12 November 2026

Put yesterday's items in a row and you can watch a complete off switch being assembled, as long as the system in question has not shipped yet.

The UK's Joint Committee on Human Rights wants a statutory body that can test an AI system and withdraw it before market release, and the government has to answer that by 14 November. New York opened a $55 million centre with more than 80 scientists to write standards and test frontier models. France got the model makers to sign the Paris Call for AI Safety and Security, with a promise to build emergency off-switches into advanced models.

Three serious efforts. All three act on the model, and all three act before deployment.

The moment that costs money sits somewhere else. It is an agent already inside your business, holding your credentials, halfway through a process nobody tested for, at twenty to four on a Tuesday. Pre-market approval does nothing in that minute. What does something is a budget ceiling that trips, a credential that expires on its own, a rollback somebody has actually rehearsed, and one named person who can pull the plug without convening anybody. None of that is in any of the three announcements.

The best case against: pre-market gates are where the leverage genuinely is. Consumer safety has worked this way for a century. You certify the brakes before the car is sold, you do not ask every driver to invent a stopping procedure in the rain. And the committee's liability shift, if it survives the government's response, pushes the cost of a failure back toward the people who built the model, which is where plenty of readers think it has always belonged.

The rest of this piece is for subscribers.

Ins7ghts Daily is free. One letter a morning from Yves Mulkers on what moved in data and AI, and why it matters to your side of the table.

Subscribe free and read on

Already subscribed? Sign in at the top of the page.

The off switch nobody is building for Tuesday

Keep Reading

Ins7ghts Daily, by Yves Mulkers