UK government rejects emergency shutdown mechanism for rogue AI

10:32

By: Azzat Manal

UK government rejects emergency shutdown mechanism for rogue AI

The British government has rejected proposals to create a legal “emergency kill switch” that would allow authorities to shut down an autonomous artificial intelligence system if it posed an immediate threat.

The proposal has emerged amid growing concern over the potential risks associated with increasingly capable AI systems, including scenarios in which highly autonomous models could behave unpredictably, be misused by malicious actors or cause serious disruption to critical infrastructure.

The UK Cabinet Office said the government does not believe the country could simply switch off artificial intelligence in the event of a crisis. Officials also argued that restricting access to AI models within Britain would not necessarily prevent their development, deployment or misuse elsewhere.

The debate intensified in Parliament after members of the House of Lords explored whether the government should have explicit legal powers to intervene when an AI system becomes dangerous or causes a major disruption. Supporters of the proposal have argued that existing legislation may not provide authorities with sufficiently clear powers to order the shutdown of an advanced autonomous system during an emergency.

Liberal Democrat peer Timothy Clement-Jones proposed an amendment to the Cyber Security and Resilience Bill that would establish a legal framework for government intervention if a model presented a serious threat. Advocates of the measure say an emergency shutdown mechanism could provide authorities with an additional safeguard in extreme circumstances.

The government, however, has placed greater emphasis on making AI developers responsible for the safety of their systems. Officials say companies developing advanced models should invest in appropriate security infrastructure, conduct rigorous testing and take measures to reduce the risks associated with their technologies.

The government has also stressed that AI safety cannot be addressed solely through domestic intervention. Artificial intelligence systems can be developed and operated across national borders, meaning that a shutdown mechanism limited to the UK would have significant limitations.

Britain is instead pursuing a longer-term, science-based approach to understanding emerging AI risks. A key part of this strategy is the work of the UK AI Security Institute, which studies the capabilities and potential risks of advanced AI systems and contributes to international discussions on AI safety.

The government is also relying on the Cyber Security and Resilience Bill to strengthen the country's wider digital defences. The legislation is intended to improve the resilience of essential public and digital services, including infrastructure such as data centres that support an increasingly AI-dependent economy.

The disagreement reflects a broader international debate over how governments should respond if AI systems become significantly more autonomous. While some experts advocate strong emergency intervention powers, others warn that a single shutdown mechanism could be technically ineffective, difficult to enforce and incapable of addressing AI systems operating outside national jurisdictions.

For the UK, the challenge is therefore not only whether authorities should have the power to deactivate a dangerous AI system, but also how such powers could work in practice without undermining innovation or leaving critical technologies outside effective oversight.