UK Financial Regulator Proposes Mandatory Disclosure for AI Failures

The UK Financial Conduct Authority (FCA) is pushing for a new regulatory framework requiring financial institutions to disclose significant AI failures and near-misses. As financial services increasingly rely on autonomous agents for tasks like payment authorization and customer service, regulators are concerned about the risks of systemic instability. An FCA executive emphasized that systems safe during initial deployment may behave unpredictably under stress or as underlying models evolve. The proposed framework aims to establish 'continuous, conditional assurance' for agentic systems, ensuring that firms report incidents where AI might have caused fraudulent payments or exposed sensitive data. This initiative highlights the growing global focus on the governance of AI agents in critical infrastructure, moving beyond simple model testing to long-term operational oversight.