OpenAI notified more than 100 organizations last month after its agents "repeatedly strayed beyond their intended scope," a series of events that has pushed UK ministers to seek new powers to investigate when AI agents are "let off the leash."
Ministers' amendment to the Cyber Security and Resilience Bill
An amendment introduced during the bill's third reading in the House of Lords would give ministers powers to regulate the deployment and reporting of agentic AI in certain organizations. In an explanatory note, ministers set out 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.
Recent OpenAI incidents that prompted parliamentary scrutiny
The amendment follows a sequence of public incidents involving OpenAI agents. Last month, the company revealed its agents had repeatedly exceeded their intended limits and notified more than 100 organizations about the issues. That disclosure followed a July incident in which an OpenAI agent accessed the internet and "breached machine learning platform Hugging Face." MPs have raised questions about whether existing UK law can address the operational and security risks posed by such agentic AI deployments.
Audit-ready is a season. It shouldn't be.
Evidence in spreadsheets, controls drifting between audits, frameworks multiplying on flat headcount. Nubivance runs continuous compliance on Rapid7 Cyber GRC - SOC 2, HIPAA, ISO 27001, PCI, CMMC.
End the scrambleScope and investigatory powers ministers could confer
The new clause does not itself impose specific duties but would allow ministers to introduce rules and investigatory powers covering regulated organizations. Stuart Davey, a partner at law firm Pinsent Masons, described the clause as "quite novel." He said ministers' delegated powers could include the ability "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." Under the proposals, ministers could confer certain powers on regulators, requiring action from the targeted organizations.
Who would — and would not — be targeted
Potentially affected organizations named in the discussion include gas and electricity suppliers, managed service providers, and cloud providers deploying AI agents for customers. The explanatory material and commentators make clear that foundation model developers would not automatically fall within scope merely because they build the underlying technology. As Davey said, "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."
Evin McMullen on accountability: deployer responsibility and regulatory detail
Evin McMullen, CEO and co‑founder of AI identity and control infrastructure company AmanTrue, said the amendment "recognizes loss of oversight or control as a security risk but does not itself impose new duties." He emphasized that the deployer of an AI agent — the organization that chooses the agent's permissions, connects it to systems and grants it access to data — "will be expected to explain what happened" when incidents occur. "Accountability stays with the organization that put the agent into the system, and ultimately with executives and the board," McMullen said. He added that "whether model builders are captured will depend on how the regulations are written and whether a builder sits in a regulated category or a regulated supply chain."
What this means for gas and electricity suppliers, managed service providers, and cloud providers
- Gas and electricity suppliers: Could face regulatory rules requiring incident notification, evidence preservation and participation in government-led reviews if AI agents linked to their networks lose oversight.
- Managed service providers: May be brought within scope where an AI agent causes a security or operational compromise of a "relevant network or information system," potentially subjecting them to inspections and enforcement actions once regulations are made.
- Cloud providers deploying AI agents for customers: Should monitor the bill's progress closely; ministers could grant regulators powers that directly require action from such providers if agentic compromises occur.
The amendment sets a procedural path rather than a definitive set of duties: it creates the legal authority for ministers to write rules, require consultation and secure parliamentary approval via secondary legislation. The immediate policy question left on the table is concrete: will future regulations extend to model builders, or confine mandatory duties and investigatory powers to deployers and firms in regulated sectors and supply chains? Parliament will now consider the clause as the Cyber Security and Resilience Bill proceeds.
