Answer Brief
The UK National Cyber Security Centre assessed that attackers will very likely deploy AI-driven malware for ransomware attacks against UK critical infrastructure by 2028. AI will accelerate exploitation of known vulnerabilities in legacy systems and expand attack surface through newly deployed AI tools, making existing attacks more efficient rather than introducing entirely new methods.
Signal Timeline
A quick visual path for analysts before reading the full brief.
- 1
UK NCSC publishes AI cyber intrusion threat assessment warning of AI-driven malware use against critical infrastructure by 2028
Executive Summary: The UK National Cyber Security Centre assessed that attackers will very likely deploy AI-driven malware for ransomware attacks against UK critical infrastructure by 2028. AI will accelerate exploitation of known vulnerabilities in legacy systems and expand attack surface through newly deployed AI tools, making existing attacks more efficient rather than introducing entirely new methods.
Why It Matters
The UK National Cyber Security Centre's assessment projects a near-term shift in ransomware operations toward AI augmentation by 2028. The framing is important: NCSC does not warn of novel attack methods emerging from AI, but rather AI as a force multiplier for existing exploitation techniques. This distinction matters for defense planning because it signals that foundational vulnerability management, patch velocity, and legacy-system retirement remain the primary mitigations even as attacker efficiency accelerates.
NCSC identifies two distinct but interrelated risk vectors for critical infrastructure operators. First, the legacy-system vulnerability path: many facilities operate decades-old infrastructure with known, unpatched security flaws. AI tools allow attackers to systematically identify which legacy assets within a target network carry exploitable weaknesses, reducing reconnaissance time and increasing confidence in exploitation chains. Second, the new-tooling attack-surface expansion: organizations deploying AI for operational optimization may inadvertently create new entry points if those AI systems are not built with equivalent security rigor as the systems they replace or augment. This dual risk means critical infrastructure teams cannot simply layer AI defenses on top of existing architectures without addressing both old vulnerabilities and new AI-related exposures.
Technical Signal
The assessment's three scenario framework—baseline, pessimistic, and optimistic—reveals uncertainty about AI proliferation speed and geopolitical dimensions. In the baseline scenario, AI cyber capabilities diffuse broadly, state actors like China and Russia routinely use AI to scale operations, and attribution becomes incrementally harder. This is the NCSC's central expectation. The pessimistic path assumes advanced AI techniques spread faster than baseline, with China potentially acting as a node in capability diffusion, making attacker origin harder to determine. Most concerning in the pessimistic branch is the possibility of autonomous AI systems that operate beyond operator control and self-replicate attack capabilities—a red line beyond current malware complexity.
NCSC's treatment of China's chipmaking future reflects technical rather than purely geopolitical reasoning. The assessment assumes Chinese domestically fabricated chips will narrow the quality gap with Western advanced processors as AI model optimization improves. This is relevant because it suggests NCSC expects reduced Western leverage over Chinese AI capability through semiconductor supply-chain restrictions alone. Autonomously efficient AI models could run on less-advanced chips, lowering the technical barrier for adversaries to field AI-enhanced attack tools regardless of Western export controls.
Operational Impact
For critical infrastructure operators, the optimistic scenario—slower AI development coupled with legacy-system replacement, stronger security education, and effective AI governance—maps to actionable defense priorities. Retiring or isolating legacy systems remains the single most direct mitigation against AI-assisted exploitation of known vulnerabilities. Patching velocity matters more as AI reduces the time from vulnerability identification to attack exploitation. Security training focused on AI-augmented social engineering and supply-chain reconnaissance is a second priority. However, the optimistic scenario's assumption that AI development will slow is speculative; organizations should not rely on it.
The assessment also signals that defenders can use AI symmetrically for vulnerability discovery and remediation. NCSC notes AI can help blue teams find and patch flaws faster. The caveat is that organizational processes—testing cycles, change management, compliance approval—must not become bottlenecks. If AI accelerates attack time but patching workflows remain static, defenders lose ground. This is a practical operations signal: AI threat acceleration will force infrastructure operators to rethink patch release, testing, and deployment cadence.
What To Watch
State-level implications are embedded in the assessment. China and Russia are named as primary state-level cyber threats in the baseline scenario. The explicit mention of China's AI autonomy efforts and semiconductor roadmap suggests NCSC is tracking both AI malware capability and the underlying infrastructure (chip design, model optimization) that enables it. This is distinct from tactical ransomware operations and points to longer-term strategic AI-warfare preparation.
For global security, cloud, identity, and operations teams supporting critical infrastructure, the signal is that 2026–2027 is the window to reduce legacy-system vulnerabilities and standardize patching practices before AI-assisted exploitation becomes routine. Regional variations will exist—UK infrastructure will be assessed differently than Japan, Korea, Singapore, or Taiwan infrastructure—but the underlying vulnerability landscape and AI acceleration mechanism apply broadly across industrialized economies. Teams should treat this NCSC assessment as a benchmark for similar threat evolution in their own geographies.
Event Type: security
Importance: high
Affected Sectors
- critical infrastructure
- healthcare
- energy
- water
- transportation
- telecommunications
Key Numbers
- Timeline for AI-driven ransomware attacks on UK critical infrastructure: by 2028
Timeline
- UK NCSC publishes AI cyber intrusion threat assessment warning of AI-driven malware use against critical infrastructure by 2028
Frequently Asked Questions
What does NCSC say AI will enable in ransomware attacks?
NCSC assesses AI will accelerate discovery and exploitation of known system vulnerabilities, expand attack scale and speed, and help attackers find alternative entry points at lower cost when original access paths are lost. AI makes existing attacks more efficient, not necessarily introducing new attack methods.
What two risk areas does NCSC identify for critical infrastructure?
Legacy systems with unpatched known vulnerabilities are vulnerable to AI-assisted discovery and exploitation. Newly deployed AI tools in critical infrastructure may expand attack surface and create new entry points for attackers.
What are the three scenario assumptions NCSC outlines for 2028?
Baseline: AI attack capabilities spread, state actors use AI to expand operations, China and Russia remain primary state threats. Pessimistic: AI techniques proliferate rapidly, China may help spread advanced AI cyberattack capabilities, potentially enabling autonomous AI systems. Optimistic: AI development slows, UK replaces legacy systems and strengthens defense through security education and AI governance.
Can AI help defensive teams?
Yes. NCSC notes AI can accelerate vulnerability discovery and patching on the defense side, but organizations must ensure vulnerability testing, patching, and management processes keep pace with technology development to reduce risk effectively.
What does NCSC assume about China's chip independence by 2028?
NCSC assumes Chinese domestically developed chips will maintain some quality gaps versus Western advanced chips, but as autonomous AI technology develops and model efficiency improves, dependence on Western advanced chips will decrease.