Back to News
PolicyAI Understanding briefing

UK cyber bill adds AI loss‑of‑control clause and frontier‑AI review

Parliament will consider two new AI provisions in the Cyber Security and Resilience Bill on 24 November, including a clause on loss of control of AI agents and a mandatory yearly review of frontier AI models.

4 min readRead the linked source
Source-provided image accompanying UK cyber bill adds AI loss‑of‑control clause and frontier‑AI review
Source referenceSource recorded
Publisher
fourweekmba.com
Source type
Linked source — primary-source status has not been established.
ContextUnderstand this in 60 seconds

Key terms

Artificial Intelligence (AI)
The broad field of building systems that perform tasks requiring pattern recognition, reasoning, language, or decision-making.
AI Agent
A software system that can observe, reason, and take actions to achieve a goal, often using tools and memory.
Test yourselfAI Ethics Quiz

What happened

On 24 November 2026, the UK House of Commons will debate the House of Lords’ amendments to the Cyber Security and Resilience Bill. The amendments introduce two AI‑related provisions that were absent from the 7 September Grand Committee version. Clause 32 empowers the Secretary of State (or an appointed inspector) to act when a cyber‑security incident is attributable to loss of oversight or control of an , with five inspection powers ranging from site entry to requiring information from abroad. Section 66 mandates a review of the regime within 12 months of the Act’s passage and thereafter at least annually, explicitly requiring consideration of “frontier artificial intelligence models and systems” and other “frontier agentic technologies”. The bill’s word count grew by roughly 5,800 words between the two versions, reflecting the added AI language.

The UK Cyber Security and Resilience Bill, originally introduced without AI language, was amended in the House of Lords on 3 November 2026. The amendments were incorporated into the Report version on 26 October 2026 and are scheduled for consideration by the House of Commons on 24 November 2026, according to Parliament’s calendar.

Clause 32 introduces an “ loss of control” trigger. It allows the Secretary of State or an appointed inspector to take five specific actions when a cyber incident is linked to an AI agent’s loss of oversight, including entering premises, seizing documents, demanding information (including from outside the UK), conducting tests or interviews, and requiring remedial steps.

Section 66 creates a statutory review cycle, requiring a review within 12 months of the Act’s commencement and thereafter at least annually. The review must assess whether additional sectors should be covered, with explicit reference to “frontier artificial intelligence models and systems” and “frontier agentic technologies”, defined as autonomous systems that match or exceed the most advanced AI models of the time.

The bill’s text expanded from roughly 37,000 words in the Grand Committee version to about 42,800 words in the Report version, a 15.8% increase, reflecting the added AI provisions and related explanatory material.

Source details: fourweekmba.com

Why it matters

The new provisions signal the UK government’s intent to treat advanced AI systems as a distinct cyber‑risk category, giving regulators concrete powers to intervene when AI agents compromise network security. By defining “frontier” AI as systems that can act autonomously and match or exceed the most advanced models of the time, the bill creates a mechanism for periodic reassessment of which sectors fall under the cyber regime, potentially expanding oversight to cloud, SaaS, and critical‑infrastructure providers that deploy cutting‑edge AI. For UK operators, the inspection powers could mean foreign‑origin data requests and on‑site audits tied to AI‑related incidents, raising compliance and data‑sovereignty considerations. AI vendors may face increased scrutiny as their models could trigger the loss‑of‑control trigger, prompting them to implement stronger governance and monitoring. Parliament’s decision will shape the regulatory landscape for AI‑driven cyber threats across the UK and could influence other jurisdictions watching the UK’s approach to AI‑specific cyber legislation.

By codifying AI‑specific cyber risk, the UK moves beyond generic cyber‑security rules, acknowledging that advanced AI can introduce novel attack vectors, such as autonomous decision‑making that bypasses human oversight.

The inspection powers give regulators a broad remit, potentially extending beyond UK borders, which could affect multinational firms that host data or AI services abroad.

The annual frontier‑AI review creates a dynamic regulatory mechanism that can adapt to rapid AI advancements, ensuring that the cyber regime remains relevant as new, more capable models emerge.

Defining “frontier” AI as matching or exceeding the most advanced models sets a high bar, meaning that today’s leading models (e.g., large language models with multimodal capabilities) will likely fall under the review, prompting operators to assess their exposure.

Interactive Mechanism

Interactive Mechanism: How It Actually Works

Explore the underlying technology behind this development interactively.

Agent Lifecycle Stage:
1
User Intent & Planning: "Audit customer refund request #4092 and settle payment."
2
Tool Calling: Emits structured JSON call crm_get_transaction(id='4092').
3
Guardrail & Verification:Paused: High-value action requires human operator sign-off.
4
Final Settlement: Refund recorded, email receipt dispatched, and audit log stored.
Core takeaway: An AI agent is not just a language model—it is a closed loop of planning, tool invocation, and environment feedback. Production systems require self-healing retries and strict human approval guardrails.
Interactive Concept Check+10 Points
AI Ethics Quiz

Why can ethical evaluation not be reduced to one model score?

What to watch next

Watch the Commons debate on 24 November for the government’s position on the two AI clauses, any amendments that may narrow or broaden the loss‑of‑control trigger, and the timeline for the first annual frontier‑AI review. Follow subsequent secondary legislation that will define “” and “loss of control”, as these definitions will determine the scope of regulator powers. Monitor industry responses, especially from UK‑based AI firms and critical‑infrastructure operators, for indications of compliance planning or legal challenges.

The outcome of the 24 November Commons debate: whether the government will endorse the clauses as‑is, propose modifications, or reject them.

The drafting of secondary legislation that will specify what constitutes an “” and the criteria for “loss of control”, which will determine the practical reach of Clause 32.

Industry feedback, especially from AI developers and critical‑infrastructure providers, which may highlight implementation challenges or legal concerns.

Future parliamentary reports stemming from Section 66’s mandated reviews, which could expand the bill’s scope to additional sectors or introduce new compliance requirements.

Related guides & quizzes

Found this useful?