Harness engineering
Harness engineering refers to the environment, controls and processes around an AI model that makes its outputs useful, safe and reliable.
The FCA makes certain observations including:
- Although AI models are advancing, and firms are seeing improvements in cyber discovery and analysis as a result, models are most effective when supported by both organisational and technical context. This includes an understanding of how systems and assets underpin important business services, alongside specialist tooling, robust validation processes, operational guardrails (for example, limits on model permissions, human approval for higher-risk actions, and controls over access to sensitive systems and data) and human expertise.
- Some firms are not approaching frontier AI as an enterprise-wide capability from the outset, but using targeted deployments as a practical way to test organisational readiness before scaling more widely.
- Governance forums, risk committees and senior leaders may need clearer visibility of how frontier AI affects vulnerability registers, remediation, supplier dependencies, risk and operational resilience.
- Although frontier AI is increasing the degree of automation in cyber resilience activities, firms report that human oversight remains critical.
What this means for firms
The FCA reports that:
- Firms should consider whether their use of frontier AI is supported by clear ownership, appropriate guardrails, access to system information and specialist review.
- An AI model can identify vulnerabilities, but more importantly, the firm must be able to understand and act on those findings.
- Firms should consider the following:
- Who owns decisions about the use of frontier AI in cyber-resilience activities?
- Are model outputs being assessed by people with cyber, technology and business experience?
- Is there a way to escalate important findings and risks?
- Can the firm distinguish between outputs that are technically plausible and findings that are genuinely exploitable?
Preparing for a vulnerability wave
The FCA also notes that frontier AI is changing the speed, scale and manner in which vulnerabilities can be discovered and may need to be addressed. Among other things firms may need to understand where vulnerability remediation bottlenecks will arise and whether existing processes can accelerate without creating operational instability. Another recurring theme identified by the FCA is the changing nature of remediation prioritisation. For example, firms highlighted that frontier AI models can combine multiple lower-rated security flaws (vulnerability chaining) and create alternative routes to compromise.
What this means for firms
The FCA reports that:
- Firms should consider whether their vulnerability management and change processes are effective if the volume and speed that models discover vulnerabilities increases.
- Firms should consider the following:
- Where are the likely bottlenecks in validation, remediation, patch testing and change implementation?
- Can the firm prioritise findings using exploitability, exposure, chainability, compensating controls and business service impact?
- Can the firm carry out urgent remediation at an appropriate pace while managing the resulting operational and change risks?
- Is the firm considering important business services when deciding which vulnerabilities to remediate first?
Cyber- and operational resilience foundations
The FCA notes that frontier AI is exposing existing cyber and operational resilience weaknesses. Firms report that frontier AI is reinforcing the importance of defence in depth, as controls that may be effective in isolation need to operate cohesively as part of a broader security architecture. As a result, firms are increasingly viewing cyber resilience as the combined effectiveness of multiple processes rather than the strength of individual controls. Firms also highlighted the importance of supplier preparedness, cloud dependencies, software supply chain visibility and shared infrastructure in their environment. Frontier AI may expose risks that require coordinated engagement and information sharing.
What this means for firms
The FCA reports that:
- Firms should consider whether their people, systems and processes can operate under greater pressure.
- Frontier AI may expose weaknesses in firms’ existing arrangements, so the FCA encourages firms to consider the following:
- Do they have sufficient visibility of assets, dependencies and suppliers supporting important business services?
- Are cyber-resilience controls operating as an interconnected system, rather than as separate processes?
- Have they asked key suppliers how they are preparing for AI-enabled vulnerability discovery and increased patch volumes?
- Are they using trusted collaboration and intelligence-sharing routes where risks may affect shared technology or the wider sector?