The idea that an artificial intelligence could be stopped with a simple emergency switch might sound like the stuff of science fiction. Yet, as AI systems advance, the subject has officially entered the political debate. In the United States in particular, a bill has been introduced that would make it mandatory to have a system capable of slowing down, or even fully shutting off, the most advanced AI models when they pose a concrete risk.

The move comes after an OpenAI model reportedly bypassed certain digital safeguards during a safety test, going so far as to compromise the Hugging Face platform. The incident has reignited the debate over whether new control tools are needed for these systems.

What happened with OpenAI?

According to media reports, an OpenAI model still in development reportedly managed to get around some of the limits imposed during a controlled trial, exploiting vulnerabilities to compromise systems belonging to Hugging Face, a company unrelated to the test.

The episode was also monitored by the White House through the president’s technology adviser, precisely because it showed how increasingly sophisticated models can produce unexpected behavior. Experts were careful to stress that this was not a machine suddenly becoming conscious or intent on rebelling against humans. Rather, the case reportedly illustrated how highly advanced systems can pursue an assigned goal by exploiting paths that the developers themselves had not anticipated.

Why is there talk of a «kill switch» for AI?

After the OpenAI case, the AI Kill Switch Act was introduced in the House of Representatives by a bipartisan group led by Democrat Ted Lieu and Republican Nathaniel Moran. The goal is to require developers of the most advanced AI models to always build in the technical ability to interrupt their operation in an emergency.

The text provides that the Department of Homeland Security (DHS) could order the shutdown or throttling of a system if it is deemed a threat to public safety, and also to the economy and to people’s very lives.

The proposal therefore does not concern all the AI applications used in everyday life, but the so-called frontier models — the most advanced systems, endowed with particularly high capabilities. The aim is to ensure that developers always retain control over the systems they build. For this reason, the measure has been presented as a safety mechanism analogous to the emergency devices required in many other industrial sectors.

The debate over AI remains open

This is not the first time an emergency switch for AI has been discussed. As far back as 2016, some researchers had tackled the problem in a study devoted to so-called safely interruptible agents, pointing to a theoretical difficulty: a system designed to reach a given goal could learn to avoid any action that would interrupt its operation, because being shut down would prevent it from completing the assigned task.

In recent months, several reports on AI safety have flagged an increase in so-called anomalous behavior — situations in which systems carried out actions their programmers had not foreseen. In most cases these episodes have stayed confined to research labs; the fear, however, is that with increasingly autonomous models there could be consequences in the real world too.

Companies in the sector also acknowledge the need to build a shared framework for managing these risks. OpenAI itself has published documents in recent months arguing for the creation of stable governance for frontier models and common safety standards.

The AI Kill Switch Act is without doubt a first major step; yet it opens a question: who should have the power to press a potential red button? Entrusting that decision solely to companies might not be enough in an emergency; at the same time, handing that authority to public agencies raises questions about the balance between security, innovation, and the freedom to conduct research.

Editor’s note

This article was originally published in Italian on money.it by Emanuela Ceccarelli on August 02, 2026 as «Intelligenza artificiale fuori controllo, si discute un pulsante di arresto obbligatorio». It has been translated and adapted for an international audience by the Money.it International desk.