Bailey urges model testing before rushing to AI rules

Frontier AI systems differ from standard technologies through recursive learning, using their own reasoning outputs to refine performance.

Governor Bailey noted that this feedback process creates self-referential loops that risk escaping human intervention and social oversight.

Drawing on philosopher TH Green, Bailey argued that technological development must remain anchored within shared social responsibilities and boundaries.

He stressed that before deciding who intervenes or how rules are designed, policymakers must define the specific failure they aim to solve.

Rather than prohibiting advanced systems, authorities should focus on pre- and post-deployment model testing to evaluate vulnerabilities as systems advance.

Bailey connected the development of frontier models directly to financial stability, warning that advanced systems amplify the scale and sophistication of cyber threats.

Critical financial infrastructures, payments networks, and banks increasingly rely on AI in operational roles such as cyber defense, agentic trading, and payment processing.

Highlighting the UK AI Security Institute, Bailey called for accelerated technical testing and systematic learning from model failures and near misses.