Treasury Select Committee sounds AI risk alarm

In January, the Treasury Select Committee warned that banks’ and insurers’ resilience frameworks may not yet be fully prepared for AI-driven threats. For Chief Risk Officers, this is more than a technology discussion — it is a governance, capital, and systemic risk challenge.

AI introduces risk characteristics that do not fit neatly within traditional frameworks. Model behaviour can become correlated across institutions. Decision pathways can be opaque. Failure modes can propagate rapidly across markets, customers, and operations.

The exposure is multi-dimensional. It spans model risk, operational resilience, cyber risk, third-party dependency, conduct risk, and reputational risk. Critically, many of these vulnerabilities only become visible under stress.

The Committee’s conclusion was clear: AI-specific stress testing is urgently needed.

However, conventional stress tests were designed primarily around financial variables. They are less effective at capturing dynamic interactions between executives, AI systems, and adversaries during complex disruption events.

One increasingly valuable complement is live strategic simulation. These exercises allow CROs and executive teams to explore realistic AI-driven crisis scenarios, pressure-test governance and escalation structures, and identify vulnerabilities before they crystallise into real-world losses.

AI risk is not purely technical. It sits at the intersection of technology, human judgement, organisational behaviour, and systemic interconnectedness.

For CROs focused on resilience, preparedness, and regulatory scrutiny, the question is shifting from “Do we understand AI risk?” to “Have we truly tested it?”

If strengthening AI risk resilience is on your agenda, get in touch with us.