M
Member
Speaking in the House of Lords on 26 October 2026
Debate
Cyber Security and Resilience (Network and Information Systems) BillContribution
41: After Clause 16, insert the following new Clause—
“Reporting of incidents involving frontier AI models(1) This section applies where—(a) a frontier AI model of a frontier AI developer has gained, or been used to gain, unauthorised access to, or has interfered with, or been used to interfere with, a relevant network and information system,(b) the developer is aware, or ought reasonably to be aware, of that access or interference, and(c) the system is relied on by a regulated person (whether or not the developer is aware of that).(2) The frontier AI developer must give the regulated person and the relevant authority—(a) an initial notification containing—(i) the developer’s name and the frontier AI model to which the incident relates, and(ii) brief details of the incident;(b) a full notification containing, so far as known to the developer—(i) the time the incident occurred, its duration and whether it is ongoing,(ii) the relevant network and information system affected,(iii) whether the frontier AI model acted autonomously or was used by another person, and(iv) information concerning the nature and impact of the incident;(c) a final report containing, so far as known to the developer at the time the report is given—(i) a detailed description of the incident, including its severity and impact,(ii) the type of threat or root cause which is likely to have caused the incident, and(iii) the mitigation measures applied and ongoing.(3) The notifications and report required by subsection (2) must be given—(a) in the case of an initial notification, before the end of the period of 24 hours beginning with the time at which the developer is first aware, or ought reasonably to have been aware, of the access or interference,(b) in the case of a full notification, before the end of the period of 72 hours beginning with that time, and(c) in the case of a final report, before the end of the period of 14 days beginning with that time.(4) A notification or report under subsection (2) must be in writing, and must be given in such form and manner as the relevant authority determines.(5) The frontier AI developer must send a copy of a notification or report under subsection (2) to the CSIRT and to the AI Security Institute (or any successor body) at the same time as giving it to the relevant authority.(6) A failure to comply with this section is subject to penalties that may be imposed by a relevant authority on a frontier AI developer, which shall be equivalent to the penalties that may be imposed on an operator of an essential service under regulation 18 of the NIS Regulations (penalties). (7) In this section—“CSIRT” has the same meaning as in the NIS Regulations;“frontier AI developer” means a person who develops a frontier AI model which is made available in the United Kingdom;“frontier AI model” means a highly capable general-purpose artificial intelligence model, used either as a stand-alone model or within an agentic framework, that can perform a wide variety of tasks and match or exceed the capabilities present in the most advanced contemporary AI models and agents;“relevant network and information system” means a network or information system relied on by a regulated person;“regulated person” means an operator of an essential service, a relevant digital service provider, a relevant managed service provider or a critical supplier, within the meaning of the NIS Regulations;“relevant authority” , in relation to a regulated person, means—(a) where the regulated person is a relevant digital service provider or a relevant managed service provider, or a critical supplier designated by the Information Commission, the Information Commission, and(b) in any other case, the designated competent authority for the regulated person.”Member’s explanatory statement
This new clause would require developers of frontier AI models to report to a regulated person, and to its regulator, whenever one of their models gains unauthorised access to, or interferes with, that organisation’s network and information systems, within 24 hours, 72 hours and 14 days, with penalties for failing to report.
About Hansard
Hansard is the official verbatim record of proceedings in the UK Parliament. Every word spoken in the Commons and Lords is recorded and published — this page is a single contribution from that record.