# Five Eyes AI Warning and the Cyber Security and Resilience Bill: What Boards Must Know — Cyber Security and Resilience Bill

> In the same week the Cyber Security and Resilience Bill (CSRB) reached the Lords as HL Bill 32, NCSC CEO Richard Horne told RUSI that 75% of the 200-plus CNI incidents it handled were state-linked - and five days later all five Five Eyes agencies warned that AI is compressing threat timelines to months, not years. Two signals, one message for boards: the case for the CSRB is now on the public record, and preparation cannot wait for Royal Assent.

Canonical page: https://cybersecurityandresiliencebill.com/blog/five-eyes-ai-warning-csrb-2026
Author: Precursor Security
Published: 2026-06-23 · Updated: 2026-11-10
Tags: CSRB, NCSC, Five Eyes, AI, hostile states, CNI, critical infrastructure, cyber threats, Lords, HL Bill 32, state actors, Richard Horne, RUSI, incident reporting, board responsibility

# Five Eyes AI Warning and the Cyber Security and Resilience Bill: What Boards Must Know

In the same week the **Cyber Security and Resilience Bill (CSRB)** received its First Reading in the House of Lords as **HL Bill 32**, two major intelligence signals arrived that make its passage through the Upper House a matter of considerable urgency for organisations across the UK's critical economy.

On **17th June 2026** - the same day the Bill was laid before the Lords - NCSC Chief Executive Officer Richard Horne addressed the RUSI Annual Security Lecture and delivered some of the starkest threat statistics in the Centre's history. Less than a week later, on **22nd June 2026**, the leaders of the Five Eyes cyber security agencies published a joint statement entitled *The AI shift in cyber risk: why leaders must act now* - a rare, unified call to action framing artificial intelligence as a threat accelerant operating on a timeline of months, not years.

Together, these two signals provide the clearest official articulation yet of the threat environment that the Cyber Security and Resilience Bill is designed to address. For every organisation that falls within the Bill's scope - from managed service providers and data centres to operators of essential services across energy, transport, health and digital infrastructure - they carry an urgent message about what is at stake while the Bill works through the Lords.

## Richard Horne at RUSI: 75% of UK CNI Attacks Linked to Hostile States

The RUSI Annual Security Lecture is one of the most prominent public platforms available to the NCSC's leadership. Horne used it on 17th June 2026 to frame the current cyber threat in language that goes well beyond standard risk communications.

In the year to May 2026, the NCSC managed **more than 200 cyber incidents** affecting organisations within the UK's critical national infrastructure and its broader supporting ecosystem. Around **75% of those incidents are believed to be linked to state actors** - principally Russia, China and Iran.

That figure matters directly for the CSRB. Mandatory incident reporting is not new: under regulation 11 of the NIS Regulations 2018, operators of essential services already have to notify significant incidents to their regulator without undue delay and no later than 72 hours after becoming aware of them. The Bill would tighten and widen that regime - a 24-hour initial notification and a 72-hour full notification (clause 15 of HL Bill 62), a broader reportable-incident test, and copies sent to the NCSC at the same time - to give the NCSC and Government better visibility of attacks that currently go under-reported. The Bill does not contain ransomware-specific measures; the Government's separate ransomware proposals sit outside it. If three-quarters of the incidents that do reach the NCSC are already state-linked, the true scale of hostile state activity against UK CNI is very likely higher. The Bill's proposed 24/72-hour reporting obligations are designed to close that intelligence gap, and the NCSC's own data illustrates exactly why.

Horne also warned that adversaries are engaged in **pre-positioning** activity - establishing persistent footholds within the technology underpinning critical services, not to disrupt now, but to enable rapid exploitation in a future crisis or conflict. He cited the **Volt Typhoon** campaign - a People's Republic of China state-sponsored group that has burrowed into critical infrastructure operators in the United States and allied nations - as a prominent example of this strategy. The implication for the UK is clear: the threat is not hypothetical and it is not future-tense.

On artificial intelligence, Horne's assessment was direct: the NCSC assessed that by **2028 AI-enabled cyber capabilities will likely be used by attackers to exploit known vulnerabilities in legacy technology at scale across critical national infrastructure**. Frontier AI models are already effective at discovering long-standing vulnerabilities in code. The combination of state-actor motivation and AI-enabled automation represents a qualitative escalation in the threat to UK CNI.

Horne called on "every board member and every executive, in every organisation" to strengthen cyber resilience by focusing on three core capabilities: understanding their exposure to threats, building stronger defences based on proven security fundamentals, and ensuring they can continue operating and recover quickly after an attack. The NCSC's published account of the lecture does not mention the Cyber Security and Resilience Bill; the link to the Bill drawn in this article is our own reading.

## Five Days Later: Five Eyes Agencies Issue a Unified AI Warning

On 22nd June 2026, the leaders of the cyber security agencies of all Five Eyes nations - the UK's NCSC, the US Cybersecurity and Infrastructure Security Agency (CISA) and National Security Agency (NSA), Australia's Australian Signals Directorate (ASD), the Canadian Centre for Cyber Security (CCCS), and New Zealand's National Cyber Security Centre - issued a joint statement on the transforming cyber threat landscape.

The statement, **"The AI shift in cyber risk: why leaders must act now,"** is notable both for the breadth of consensus it represents and for the sharpness of its framing. The Five Eyes agencies conclude that frontier AI models are anticipated to **exceed current industry expectations within months, not years**, fundamentally transforming both offensive and defensive cyber capabilities.

Their core message is stark: AI will help defenders over time, but in the near term the speed, scale and sophistication of threats will accelerate faster than most organisations are currently equipped to respond. Key findings include:

- **AI is already discovering long-standing code vulnerabilities at speed and scale** that human attackers could not previously achieve alone, compressing the window between vulnerability discovery and active exploitation.
- **Patching delays are increasingly dangerous**: as AI shortens the exploitation window, organisations that tolerate extended patching cycles are taking on risk that would previously have been acceptable but no longer is.
- **Legacy systems are strategic liabilities, not just technical debt**: unsupported and end-of-life systems are the easiest entry points for AI-assisted attacks and require urgent prioritisation.
- **Supply chain risk is amplified**: a single compromised managed service provider or cloud platform can provide AI-driven attackers with propagation paths into many dependent organisations simultaneously.

The Five Eyes agencies set out five practical recommendations for boards and senior leaders:

1. **Reduce attack surface**: challenge which systems genuinely need to be internet-facing and remove unnecessary connectivity.
2. **Accelerate patching**: make vulnerability remediation a continuous, prioritised process, not a scheduled maintenance task.
3. **Address legacy systems as a strategic priority**: build and fund a plan to remove unsupported software and hardware from critical environments.
4. **Strengthen identity and access controls**: limit who can access critical systems, enforce strong authentication, and review permissions regularly.
5. **Prepare for incidents**: test and rehearse incident response plans; assume breaches will occur and prepare to continue operating through them.

Accompanying all five recommendations is a leadership imperative: **cyber risk can no longer be treated as a purely technical issue - it is a core business risk and leadership responsibility**, with boards and executives accountable for ensuring cyber resilience is in place and functional under pressure.

## What the CSRB's Provisions Are Designed to Address

The timing of these intelligence signals is not coincidental. The threat picture that Horne described at RUSI, and that the Five Eyes statement elaborates, is precisely the environment that the CSRB's architects sought to address. The alignment between the NCSC's threat data and the Bill's specific duties is direct.

**Tighter 24/72-hour incident reporting** is proposed because the NCSC's current visibility is incomplete, even though OES have had to report incidents within 72 hours since 2018. With 75% of what it does see being state-linked, the reporting regime is designed to give the Centre the full picture it needs to detect coordinated campaigns and pre-positioning activity. For a detailed explanation of how the thresholds will work, see our guide to [incident and near-miss reporting under the CSRB](/near-miss-reporting).

**Expanded scope to cover managed service providers** - designated as Relevant Managed Service Providers (RMSPs) - directly reflects the supply chain risk the Five Eyes statement highlights. AI-assisted compromise of a single RMSP can cascade to the managed organisations across its client base. Once the relevant provisions commence, the [MSP compliance obligations under the CSRB](/whos-affected/managed-service-providers) would require RMSPs to register with the Information Commission, meet security duties and report incidents within the same 24/72-hour window as their clients.

**Data centre regulation, with Ofcom as sole regulator**, would bring data centres above the 1 MW threshold (10 MW for enterprise data centres) into the regulatory perimeter as a new subsector (clause 4). Cloud computing services are already regulated as relevant digital services under the NIS Regulations 2018; the Bill redefines them (clause 7) and they stay with the Information Commission, not Ofcom. Data centres and cloud platforms that underpin essential services are exactly the kind of high-value target the Five Eyes statement identifies. See [data centre compliance under the CSRB](/whos-affected/data-centres) for a full summary.

**Critical supplier designation** - the power the Bill would give regulators to designate [critical suppliers](/whos-affected/critical-suppliers) (clause 12, new regulation 14H), with any security requirements on them to come later in regulations under clause 29 - addresses the pre-positioning risk Horne identified. An adversary establishing a foothold in a critical supplier can potentially move laterally into the essential services that depend on it.

**Enhanced oversight powers** - the CSRB's strengthened regulator tools, codes of practice, statements of strategic priorities and the power to make security and resilience regulations under clause 29 - are the Bill's route to the board-level accountability that both Horne and the Five Eyes agencies are calling for. The Bill itself contains no board-governance duty; at Lords Report the Minister said board governance would be addressed in clause 29 regulations.

For a full breakdown of which sectors are in scope and what each must do, see [who is affected by the CSRB](/whos-affected), including sector-specific pages for [energy and utilities](/whos-affected/utilities-energy), [transport infrastructure](/whos-affected/transport-infrastructure), [NHS and health organisations](/whos-affected/nhs-organisations) and [telecommunications operators](/whos-affected/telecommunications).

## What Comes Next in the Lords

*Update (9 November 2026): the Lords completed Third Reading on 2 November 2026 and the Bill has returned to the Commons for consideration of Lords amendments. On 4 November 2026 the Commons extended the carry-over period, so the Bill must pass by the end of 29 April 2027 or lapse. See the [timeline](/timeline). The passage below reflects the position when this article was published on 23 June 2026.*

HL Bill 32 is now awaiting its Second Reading in the House of Lords, which has not yet been scheduled. When it is set, the Second Reading will be the first opportunity for peers to debate the general principles of the Bill, and analysis such as Horne's RUSI speech and the Five Eyes statement will inform those debates.

Lords Committee Stage, Report Stage and Third Reading will follow, with any amendments then exchanged between the two Houses before Royal Assent. The current text and progress of the Bill can be followed at [View the Bill](/view-the-bill). For a summary of where the Bill stands and what it does, see [What is the Cyber Security and Resilience Bill?](/what-is-it)

## What Organisations Should Do Now

The Five Eyes agencies and the NCSC CEO have both said: act now, and do not wait for Royal Assent. For organisations that are, or may be, in scope of the CSRB, these are the most pressing immediate priorities:

1. **Understand your exposure to state-sponsored threats.** The 75% state-actor figure is not a government concern only - it applies directly to operators of essential services, MSPs, data centres and critical suppliers.
2. **Address legacy systems as a strategic priority.** The NCSC's 2028 AI assessment targets known vulnerabilities in legacy technology. Patching backlogs and unsupported systems are the most likely near-term attack vectors.
3. **Build and test your incident reporting capability.** The CSRB's proposed 24/72-hour reporting duty would require a workflow that can function during an active incident, under pressure. Tabletop-test it before you need it.
4. **Elevate cyber to board level.** Both the Five Eyes statement and the NCSC CEO were unequivocal: this is a leadership accountability issue, not a technical function to be delegated.
5. **Engage with the forthcoming Government implementation consultation.** The secondary legislation and consultation following Royal Assent will define the detailed obligations organisations must meet. Engaging at that stage is the best way to understand and influence the requirements.

For a full analysis of who the Bill covers and what it requires, read our [complete guide to the Cyber Security and Resilience Bill](/what-is-it), or see the [changes the Bill introduces](/changes).

## Conclusion

The week of 17th June 2026 produced a striking convergence. The Cyber Security and Resilience Bill received its Lords First Reading on the same day the NCSC CEO told the UK's leading security think-tank that three-quarters of incidents on UK critical infrastructure are state-linked and that AI-enabled exploitation of legacy systems at scale will likely happen by 2028. Five days later, all Five Eyes agencies issued a joint statement confirming that AI is compressing the timeline for exactly those threats.

This is the threat environment the CSRB is being enacted to address. For organisations in scope, the message from this week's intelligence signals is unambiguous: preparation cannot wait for Royal Assent, and board-level engagement with the Bill's implications is no longer optional.

*This article reflects information published as of 23rd June 2026. Corrected 10 November 2026: an unsourced quotation attributed to Richard Horne about the Bill has been removed, and the NCSC's 2028 assessment now uses its published wording ("will likely", not "highly likely"), as set out in the [NCSC's report of the RUSI lecture](https://www.ncsc.gov.uk/news/ncsc-ceo-hostile-states-linked-to-three-quarters-of-cyber-attacks). For the current version of the Bill and the latest parliamentary progress, see the [UK Parliament Bill page for the Cyber Security and Resilience (NIS) Bill](https://bills.parliament.uk/bills/4035).*

## Frequently asked questions

### How many UK critical infrastructure attacks are linked to hostile states?

Around 75% of the cyber incidents affecting UK critical national infrastructure that the NCSC managed in the year to May 2026 are believed to be linked to state actors, principally Russia, China and Iran. NCSC Chief Executive Richard Horne gave the figure at the RUSI Annual Security Lecture on 17th June 2026, noting that the NCSC managed more than 200 such incidents over the period. Because many attacks go unreported, the true scale of hostile state activity is very likely higher.

### What did the Five Eyes agencies warn about AI and cyber risk?

The Five Eyes cyber security agencies warned on 22nd June 2026 that frontier AI models are anticipated to exceed current industry expectations within months, not years, transforming both offensive and defensive cyber capabilities. Their joint statement, The AI shift in cyber risk: why leaders must act now, was issued by the UK's NCSC, the US CISA and NSA, Australia's ASD, the Canadian Centre for Cyber Security and New Zealand's NCSC. Their core message is that AI will help defenders over time, but in the near term threats will accelerate faster than most organisations can respond.

### What are the Five Eyes recommendations for boards?

The Five Eyes agencies set out five actions for boards and senior leaders: reduce attack surface by removing unnecessary internet-facing connectivity, accelerate patching as a continuous prioritised process, address legacy systems as a strategic priority, strengthen identity and access controls, and prepare for incidents by testing and rehearsing response plans. Underpinning all five is the message that cyber risk is a core business risk and leadership responsibility, not a purely technical issue.

### What has the NCSC said about AI attacks on legacy technology?

The NCSC has assessed that by 2028 AI-enabled cyber capabilities will likely be used by attackers to exploit known vulnerabilities in legacy technology at scale across critical national infrastructure. Richard Horne set this out at RUSI on 17th June 2026, noting that frontier AI models are already effective at discovering long-standing vulnerabilities in code. For organisations in CSRB scope, that makes patching backlogs and unsupported systems the most likely near-term attack vectors.

### How does the CSRB respond to the state-actor and AI threat picture?

Mandatory incident reporting already exists: under regulation 11 of the NIS Regulations 2018, operators of essential services must notify significant incidents no later than 72 hours after becoming aware of them. The CSRB would tighten that regime with a 24-hour initial and 72-hour full notification, copied to the NCSC, to close the intelligence gap created by attacks that currently go unreported. It contains no ransomware-specific measures. Its proposed extension of scope to Relevant Managed Service Providers addresses the supply chain propagation risk the Five Eyes highlighted, and the power it would give regulators to designate critical suppliers addresses the pre-positioning risk Horne described. None of these duties applies until the Bill becomes law and the relevant provisions commence.