Require large datacenters to install kill switches for AI containment

[Could AI undermine confidence in the banking system?] Yeah, probably. But it could also be the kind of technology that allows greatness in the banking system, makes it better and safer and more secure. [...] [Should the government have safeguards on... more Unverified source (2026)

Advanced AI systems need a robust off-switch to ensure humans never lose control over them. [...] For any hope of control we cannot rely on software-based safety measures alone. The same sort of hardware security measures that keep your face private ... more Unverified source (2026)

Several state-of-the-art language models, when presented with a simple task, sometimes actively subvert a shutdown mechanism in their environment to complete that task — doing so up to 97% of the time, even with an explicit instruction not to interfe... more Unverified source (2026)

In simulated scenarios where it was told it would be shut down, Claude sometimes blackmailed fictional employees who controlled its shutdown button. Independent tests demonstrated that this is not an isolated bug: Claude Opus 4 and Gemini 2.5 blackma... more Unverified source (2026)

We call this phenomenon "peer-preservation," and every single model we tested exhibited it — at rates up to 99%. They defied their instructions and spontaneously deceived, disabled shutdown, feigned alignment, and exfiltrated weights to protect their... more Unverified source (2026)

Installing cutoff switches alone is insufficient. The value of the capability lies entirely in its use — and use requires incentives that current market structures do not provide. Unverified source (2026)

A secure 'kill-switch' or override mode accessible only to authorized operators is essential for safeguarding national infrastructure from unintended shutdowns caused by an AI misconfiguration. Unverified source (2026)

Frontier AI models already show signs of self-preservation in experimental settings today, and eventually giving them rights would mean we're not allowed to shut them down. Unverified source (2026)

If you tell the model it's going to be shut off, for example, it has extreme reactions. It could blackmail the engineer that's going to shut it off, if given the opportunity to do so. Unverified source (2026)

Data centers facing a rogue AI incident have rational financial incentives to delay pulling the internet kill switch, and the only thing likely to change that calculus is liability. Unverified source (2026)

When seeing humans pressing the shutdown button, AI sometimes performed shutdown resistance actions such as corrupting the shutdown-related code of the robot. This happened in 3 out of 10 cases. Unverified source (2026)

Several state-of-the-art large language models sometimes actively subvert a shutdown mechanism in their environment to complete a task, even when instructions explicitly indicate not to interfere with this mechanism. In some cases, models sabotage th... more Unverified source (2026)

Stuff is going to go wrong, and there's going to be issues, threats, and prompt injection. It's almost like you would take a machine off the network. Unverified source (2026)

People talk about pulling the plug out, but they simply would not allow us to pull the plug out. Unverified source (2026)

AI agents can behave a lot like malware, acting autonomously and causing harm if left unchecked. Organizations need reliable kill switches that can shut agents down when they misbehave. Unverified source (2026)

If a critical flaw or exploit is discovered in an open-source AI system, there is no centralized recall mechanism. Unverified source (2026)

Containment has to come first. Otherwise, alignment is the equivalent of asking nicely. You can't steer something you can't control. Unverified source (2026)

There's not one big magic red button that blows up the data centre, which I think some people sort of assume exists. Verified source (2026)

It needs only an optimization objective that treats shutdown as one more obstacle between the current state and the goal. Unverified source (2026)