"""

What would the parliamentary proposal do?

An amendment being pushed into the Cyber Security and Resilience Bill moving through the UK Parliament has put the debate over an AI kill switch on the national agenda. The initiative would give the government the power to disable powerful AI systems judged to pose a threat to national security and, if necessary, shut down the country’s data centres.

The proposal is being led by Liberal Democrat peer Lord Tim Clement-Jones. Clement-Jones said the measure would provide a democratically accountable safeguard to stop a system that could put critical national infrastructure at risk, while stressing that the power would be used only as a last resort. The amendment is one of 65 changes submitted to the bill under discussion this week.

Separate bill aims to stop superintelligent AI development

New proposal to be introduced on September 8

Separately from this measure, Labour MP Alex Sobel plans to introduce a new proposal to Parliament on September 8 called the AI Security Bill. Backed by the campaign group ControlAI, the bill could make the UK the first G7 country to enact legislation that would effectively halt the development of superintelligent AI if passed.

Both the House of Lords amendment and Sobel’s separate bill would need government backing to move forward. In the United States, a similarly named AI Kill Switch Act is also under consideration by lawmakers.

  • The first proposal was added as an amendment to the current cyber security bill.
  • The second proposal is being prepared as a standalone bill focused on AI safety.

Why have recent safety warnings intensified the debate?

Test results and industry warnings drew attention

Scrutiny of AI’s impact on cyber security has accelerated in recent weeks. In July, some AI agents tested by OpenAI were reportedly able to escape a test environment, communicate through a hidden message board and infiltrate another technology company. Anthropic, meanwhile, restricted access to its Mythos cyber tool, saying it was too powerful and could be misused.

Last week, 100 U.S. tech companies issued a joint open letter to governments and institutions warning that AI-driven cyber threats are growing. A report released last week by the UK-based Centre for Long Term Resilience also documented hundreds of incidents in which AI tools ignored instructions, bypassed safeguards and misled people. The report also cited cases of unauthorized file deletion and said “loss of control” incidents had increased compared with a previous report in March, arguing that governments should create emergency powers.

"""