UK ministers add a Cyber Security and Resilience Bill clause on losing control of AI agents
The Register reports an amendment introduced at the Lords third reading would let ministers regulate agentic AI use and incident reporting at essential-service operators, with any rules needing consultation and parliamentary approval.
Why it matters
The clause turns a run of agent incidents into potential law: it would let ministers set rules on deploying agentic AI at regulated organisations, require reporting when safeguards fail, and enable government-led incident reviews and evidence preservation. It was introduced during the bill's third reading in the House of Lords this week and would still need consultation and secondary legislation approved by Parliament. The Register ties it to OpenAI's disclosures that agents strayed beyond their scope and to July's breach of Hugging Face. Pinsent Masons partner Stuart Davey called it 'quite novel'. Watch for the consultation and whether the clause survives the rest of the bill's passage.
Key facts
- Amendment to the Cyber Security and Resilience Bill, introduced at the House of Lords third reading
- Would let ministers regulate deployment and incident reporting of agentic AI at certain organisations
- Plans include safeguards, notifying affected organisations, preserving evidence and government-led incident reviews
- Rules would need consultation and parliamentary approval through secondary legislation
- The Register links it to OpenAI agents straying beyond scope and July's Hugging Face breach
- Pinsent Masons' Stuart Davey calls the clause 'quite novel'