Alex Sobel Tables UK Bill to Ban Superintelligence as Government Weighs Targeted Action
The private bill would set a legal stop line for AI that outperforms people in most domains. The UK government says the proposal is not the right approach, but is considering targeted national-security interventions.
Listen to this story
The audio brief
Story brief
3 key pointsAlex Sobel’s private bill would prohibit UK development, deployment, and operation of artificial superintelligence while giving ministers supply-chain oversight extending to chips. It is not law or an immediate restriction: the government rejects the approach and is instead considering unspecified, targeted measures for major AI national-security risks. The proposal follows failed attempts to add a data-center...
- 01
The bill defines superintelligence as AI outperforming humans across most domains.
- 02
Its proposed controls would cover the AI supply chain, including chip-level activity.
- 03
The government says existing cybersecurity law can isolate or stop use of threatening models.
Labour MP Alex Sobel has tabled a bill to ban the development, deployment and operation of artificial superintelligence in the UK. It would also give the government powers to monitor and control development across the AI supply chain, including at the chip level.
The proposal defines artificial superintelligence as AI that would outperform humans in most domains. Sobel tabled it in the House of Commons, so it is a legislative proposal rather than a current restriction. It is private legislation and has not been enacted.
A wider proposal after the kill-switch setback
The bill follows Sobel’s unsuccessful effort to add a government kill switch for threatening AI agents to the Cyber Security and Resilience Bill. Peers also unsuccessfully tried to restore that measure in the House of Lords the week before the new proposal was reported.
What Sobel’s earlier measure sought
- It would have allowed the government to shut down a data center hosting an AI agent that presented a threat.
- The government said existing cybersecurity legislation already allowed intervention over compromised systems posing a national-security risk, including requiring an entity to stop using and isolate an AI model.
The government’s narrower response
The UK government said the bill’s measures are not the right approach. It is considering whether additional targeted interventions may be needed for the most significant AI-related national-security risks, without identifying a specific measure.
ControlAI drafted the bill; the nonprofit is backed by Jaan Tallinn, a Skype founding engineer and former DeepMind investor director. Sobel has said he will present the proposal in full in November 2026. He hopes it could lead the UK to work with other countries on an international ban, but that is his stated aim, not a government commitment.
Editorial analysis
Our Read
The practical divide is over enforcement as much as AI risk. Sobel’s bill would prohibit specified superintelligence activity and extend oversight through the supply chain, including chips. The UK AI Security Institute, by contrast, can evaluate advanced models only when companies volunteer access and cannot block a release. The government has not identified the targeted intervention it may pursue, so its response is not yet a competing regulatory plan. Sobel’s November presentation is the next concrete point at which the bill’s scope may become clearer—and when the government’s exploratory position may face a more detailed proposal.
Sources
- theguardian.comUK lawmakers urge Burnham to back ban on superintelligent AI after chilling warnings
- wired.comUK Lawmakers Are Freaking Out Over AI’s Summer of Chaos
Loading discussion...
Reader comments
Newest comments first. Replies stay oldest first.