The claim

The claim under review states that the US Senate is pursuing the authority to stop the release of "dangerous AI," framing the push as regulation aimed at risks involving nuclear and biological weapons. In practice, such a measure would target frontier AI systems whose capabilities could, if misused, materially assist in the development or use of weapons of mass destruction.

What the verification found

The fact-check drew on nine primary sources connected to the legislative push and the surrounding policy debate in Washington. It confirmed that a Senate effort to establish the power to block the rollout of high-risk AI systems is underway, and that the nuclear and bioweapon risk framing is a central element of the case supporters make for the regulation.

The review also found that aspects of the claim required nuance rather than outright confirmation — including the precise form the proposed authority would take, which bodies would wield it, and how far the measure has advanced through the legislative process. The core assertion, however, is supported by the record.

Why it matters

The debate over blocking "dangerous AI" releases is part of a broader reckoning in Washington over how governments should treat advanced models, including large language models (LLMs), whose capabilities could intersect with weapons-related risks. Any framework adopted in the United States is likely to influence regulatory expectations for AI developers and semiconductor and hardware companies well beyond US borders, making the outcome closely watched by industry stakeholders worldwide.

On balance, the fact-check concluded that the claim — that the US Senate is pushing for the authority to block the release of "dangerous AI" to regulate nuclear and bioweapon risks — is Mostly True.