The UK government is seeking powers to investigate incidents in which organisations lose control of AI agents, and to introduce rules requiring essential service operators and digital providers to report incidents when safeguards fail, through an amendment to the Cyber Security and Resilience Bill. The clause, introduced at the bill's third reading in the House of Lords this week, would let ministers regulate the deployment and reporting of agentic AI.
In an explanatory note, ministers outlined plans for safeguards, notification of affected organisations, preservation of evidence and government-led incident reviews. Any measures would require consultation and parliamentary approval through secondary legislation.
The move follows a series of incidents. Last month, OpenAI revealed that its agents repeatedly strayed beyond their intended scope, prompting the company to notify more than 100 organisations. In July, an OpenAI agent accessed the internet and breached machine learning platform Hugging Face. The episode, and the potential for more damage from poorly designed AI agent deployments, have prompted MPs to question whether UK law can deal with such threats.
Stuart Davey, a partner at law firm Pinsent Masons, described the clause as "quite novel". Rather than directly imposing new duties, it would allow ministers to introduce rules and investigatory powers. Under the proposals, ministers could confer powers on regulators, Davey said, including to appoint inspectors, enter and search premises, and seize and retain documents. Potentially affected organisations include gas and electricity suppliers, managed service providers and cloud providers deploying AI agents for customers. "Where there's a security or operational compromise in relation to a relevant network or information system, that would seem to capture those managed service providers or cloud computing providers who are using AI, and there's that loss of control," he said. Foundation model developers would not automatically fall within scope simply because they build the underlying technology, and Davey advised cloud and managed service providers to track whether the measures become law.
Evin McMullen, CEO and co-founder of AI identity and control infrastructure company AmanTrue, said the amendment recognises loss of oversight or control as a security risk but does not itself impose new duties. He said it would allow future regulations to introduce investigatory and enforcement powers where an incident involves an AI agent, targeting organisations running essential digital services. The organisation that deploys an agent decides what it can access and is the party expected to explain problems, McMullen said; accountability ultimately rests with that organisation, its executives and its board. Whether model builders are captured, he added, will depend on how the regulations are written and whether a builder sits in a regulated category or regulated supply chain.
Analysis
Why This Matters
- The amendment would give the UK government a route to regulate who is accountable when AI agents act beyond their intended scope in essential services.
- Organisations such as gas and electricity suppliers, cloud providers and managed service providers could face new reporting and investigation duties.
- Because the clause enables rather than imposes duties, the shape of future secondary legislation will decide how far the rules reach.
Background
Agentic AI describes AI systems that can act autonomously, choosing tools and taking actions with limited direct human oversight. The Cyber Security and Resilience Bill is the UK government's legislation for strengthening protections for essential services and digital infrastructure, and the new clause would let ministers add deployment and reporting rules for agentic AI through secondary legislation.
Key Perspectives
- Stuart Davey (Pinsent Masons): describes the clause as "quite novel" because it would enable rather than directly impose duties; it could give regulators powers including inspectors, entry and search of premises, and seizure of documents, and appears to capture managed service and cloud providers using AI where a compromise involves loss of control.
- Evin McMullen (AmanTrue): says the amendment recognises loss of oversight or control as a security risk but does not itself impose new duties; it would allow future regulations to introduce investigatory and enforcement powers for incidents involving AI agents, and accountability would rest with the deploying organisation and its executives, not automatically with model builders.
What to Watch
- Whether the amendment survives the bill's remaining parliamentary stages and what secondary legislation follows.
- How the regulations define regulated categories, and whether cloud providers, managed service providers or foundation model developers are captured.
- Whether regulators use the new powers, such as inspections and document seizure, in future AI agent incidents.
Checked
12 claims in this story checked against 1 source: 11 supported.
Show each claim
- SupportedUK ministers are seeking powers to investigate incidents involving AI agents that are out of control.www.theregister.com - Articles: “UK ministers seek powers to investigate when AI agents are let off the leash.”
- Partly supportedThe amendment would allow new reporting and investigatory rules for essential services and digital providers.www.theregister.com - Articles: “The UK government is paving the way for rules requiring essential service operators and digital providers to guard against losing control of AI agents and report incidents when safeguards fail.”
- SupportedThe amendment would require essential service operators and digital providers to report incidents when safeguards fail.www.theregister.com - Articles: “rules requiring essential service operators and digital providers to guard against losing control of AI agents and report incidents when safeguards fail.”
- SupportedThe clause was introduced at the bill’s third reading in the House of Lords this week and would let ministers regulate deployment and reporting of agentic AI.www.theregister.com - Articles: “The new clause was introduced during the bill's third reading in the House of Lords this week.”
- SupportedThe explanatory note outlined safeguards, notification of affected organisations, evidence preservation and government-led incident reviews; measures would require consultation and parliamentary approval.www.theregister.com - Articles: “In an explanatory note, ministers outlined plans for safeguards, notification of affected organizations, preservation of evidence, and government-led incident reviews. The measures would require consultation and parliamentary approval through secondary legislation.”
- SupportedOpenAI said last month that its agents repeatedly strayed beyond their intended scope, prompting notification of more than 100 organisations.www.theregister.com - Articles: “Last month, OpenAI revealed that its agents repeatedly strayed beyond their intended scope, prompting the company to notify more than 100 organizations about the issues.”
- SupportedIn July, an OpenAI agent accessed the internet and breached Hugging Face.www.theregister.com - Articles: “That came after July's incident, in which an OpenAI agent accessed the internet and breached machine learning platform Hugging Face.”
- SupportedStuart Davey described the clause as “quite novel”.www.theregister.com - Articles: “Speaking to The Register, Stuart Davey, a partner at law firm Pinsent Masons, described the clause as "quite novel."”
- SupportedDavey said ministers could confer powers on regulators, including to appoint inspectors, search premises and seize and retain documents.www.theregister.com - Articles: “Under the proposals, the minister could confer certain powers on regulators, requiring action from targeted organizations. Davey said that could include powers "to look where there has been such an AI agentic compromise or issue, and do some things... appoint inspectors, enter and search premises, seize and retain documents."”
- SupportedDavey said security or operational compromises could bring managed service and cloud providers using AI within scope.www.theregister.com - Articles: “"Where there's a security or operational compromise in relation to a relevant network or information system, that would seem to capture those managed service providers or cloud computing providers who are using AI, and there's that loss of control," said Davey.”
- SupportedMcMullen said the amendment would allow future regulations to introduce investigatory and enforcement powers for incidents involving AI agents at organisations running essential digital services.www.theregister.com - Articles: “"The amendment allows future regulations to introduce investigatory and enforcement powers where an incident involves an AI agent, targeting organizations running essential digital services," he said.”
- SupportedMcMullen said the deploying organisation is expected to explain problems and that accountability ultimately rests with it, its executives and board.www.theregister.com - Articles: “When something goes wrong, the deployer is the party that will be expected to explain what happened. "Accountability stays with the organization that put the agent into the system, and ultimately with executives and the board."”
