The six agencies behind the Five Eyes partnership have issued a joint statement urging executives to treat AI-driven cyber risk as a core business problem — and to fix the fundamentals before frontier models close the gap between a vulnerability and its exploitation.
Summary
- The heads of the Australian, Canadian, New Zealand, UK and US cyber agencies (ACSC, CCCS, NCSC NZ, NCSC UK, CISA and NSA) released a joint “call to action” on 22 June 2026 on AI and cyber risk.
- Their central warning: frontier AI is reshaping both offence and defence, and the relevant planning horizon is now measured in months rather than years.
- AI is shrinking the window between vulnerability discovery and exploitation, which puts a premium on fast patching and a smaller attack surface.
- The agencies frame cyber resilience as a board-level business risk, not an IT problem — and list five “not new, but now urgent” actions for leaders.
- Nothing in the statement is novel security advice; the news is the unified, executive-aimed urgency from all six agencies at once.
What the agencies actually said
This is a joint statement, not a technical advisory. It is signed by the heads of all six agencies that make up the Five Eyes cyber security partnership, and it is pitched squarely at boards and executives rather than at SOC analysts.
The framing is that AI cuts both ways. It will help defenders over time, but in the near term it accelerates the speed, scale and sophistication of attacks. The agencies are blunt about the timeline, stating that the shift in offensive and defensive capability is coming in “months, not years” as frontier models exceed current industry expectations.
The practical consequence they highlight is timing. AI lowers the barrier to entry for malicious actors and compresses the gap between a vulnerability being discovered and being exploited. That single dynamic drives most of the advice that follows.
The core principles
The agencies anchor the statement on three principles that will be familiar to anyone in security leadership:
- Secure-by-design and secure-by-default should be standard practice, not an aspiration — and that expectation is explicitly extended to vendors.
- Defence in depth still holds. No single tool or technology should be load-bearing for your resilience.
- New vulnerabilities, including zero-days, will keep emerging as AI systems evolve, so preparedness matters more than prevention alone.
The blunt corollary is that breaches will happen. The agencies’ position is that the job of leadership is to contain them fast and stop them escalating into operational and financial crises.
Five actions for leaders
The statement lists five practical actions and is candid that none of them are new — the change is the urgency:
- Reduce your attack surface. Challenge whether systems need external exposure at all, and isolate the ones that do not.
- Accelerate patching. Because AI is shortening the discovery-to-exploitation window, slow patch cycles now carry more risk — especially for operational systems with long maintenance windows.
- Address legacy systems. Unsupported kit is framed not as technical debt but as a strategic liability.
- Strengthen identity and access controls. Limit access to critical systems, enforce strong authentication, and review permissions regularly.
- Prepare for incidents before they happen. Test response plans, train teams, and assume breaches will occur — with the emphasis on fast containment and recovery.
The agencies also push leaders to use AI on the defensive side: earlier vulnerability detection, better software quality, anomaly monitoring and faster incident response. Their line is that success comes from getting the basics right and acting quickly, not from owning the most tools.
Why it matters for CISOs
For most security leaders, the value here is not the content — it’s the cover. Every item on this list is already on your roadmap, and probably already on a risk register somewhere. What you now have is a unified statement from six national agencies, aimed directly at boards, that reframes patch latency, legacy systems and identity hygiene as core business risk rather than IT housekeeping.
That changes the conversation you can have upstairs. The “timeline is months, not years” framing is a useful lever for accelerating funding and decision-making on programs that have stalled on cost or change-freeze grounds — particularly legacy decommissioning and patch-cycle reform for operational systems, which are usually the hardest sells.
There is also a quieter signal worth reading. The explicit call-out of vendors and secure-by-design suggests continued regulatory and procurement pressure on suppliers. If your third-party risk and procurement security requirements are soft, this is a reasonable moment to tighten them while the wind is at your back.
One caveat for your own messaging: the statement leans on the “months, not years” claim without putting hard evidence behind it. It is a directional warning from agencies with visibility you don’t have, so it’s fair to use — but treat it as a planning prompt, not a forecast you can defend with data.
What to do now
- Map this statement to your existing program and identify which of the five actions are weakest, then use the joint statement as the supporting reference in your next board or risk-committee update.
- Pull your real patch-latency numbers for internet-facing and operational systems. If the discovery-to-exploitation window is shrinking, your current SLAs are the thing most exposed by this warning.
- Re-run an attack-surface review focused on unnecessary external exposure, and flag any legacy or unsupported systems as strategic risk items rather than deferred maintenance.
- Check identity and access hygiene on critical systems — authentication strength and stale permissions — since this is the control set most often eroded by drift.
- Confirm your incident response plan has actually been tested, not just written, with containment and recovery as the measured outcomes.
- Revisit vendor and procurement security requirements against the secure-by-design expectation, especially for suppliers introducing AI features into your environment.
