Risk & governance
AI alignment
Making AI behaviour consistent with intended goals and constraints.
Definition
Alignment research and practice seek to ensure models and agents pursue specified objectives without harmful or unintended behaviour.
Why it matters
General provider alignment does not encode a bank's products, policies, conduct duties, risk appetite or approval authorities.
Related concepts
- Constitutional AI
Training and steering models using an explicit set of written principles.
- Red teaming
Adversarial testing designed to uncover failures and unsafe behaviour.
- Guardrail
A technical or procedural control that constrains system behaviour.