1. Supporting Consumers in Vulnerable Circumstances
How can AI help identify and support consumers in vulnerable circumstances, without creating new forms of exclusion?
Vulnerability is dynamic, and the point at which a consumer most needs support is often the point at which they are least able to ask for it. AI could identify difficulty earlier and adapt how firms communicate. It could equally turn inference about circumstance
into a basis for exclusion, or replace human contact when it matters most.
2. Fairness and Explainability in Automated Decisions
How can firms demonstrate that AI-driven decisions in financial services are fair and explainable and can be understood?
Creditworthiness assessment and life and health insurance pricing are high-risk uses under the AI Act. The Bank has been here before. The differential pricing review led to the prohibition on price walking in 2022. AI-driven personalisation raises the
same question continuously, at speed, and with far less visibility.
3. Strengthening Consumer Resilience to Fraud and Scams
How can defences against fraud be strengthened as AI raises the sophistication of attacks?
Firm-side controls have improved considerably. The position of the consumer has not. A person can be persuaded to authorise a payment, coached to defeat a firm's own warnings, and convinced a message comes from their provider. A control the consumer can
be talked around is not a control.
4. Accountability as AI Moves from Advising to Acting
Where does accountability sit as AI moves from advising to acting?
The line between a system that informs, one that gives regulated advice, and one that executes is becoming hard to identify. Agentic systems can initiate payments and delegate to other agents with limited oversight at the point of action, often outside
any regulated relationship with the consumer.