Trustworthy AI

Trustworthy AI systems exhibit seven key characteristics that must be balanced based on context:

1. Valid and Reliable: Systems perform accurately and consistently under expected conditions
2. Safe: Systems do not endanger human life, health, property, or environment
3. Secure and Resilient: Systems maintain confidentiality, integrity, and availability while withstanding adverse events
4. Accountable and Transparent: Systems provide appropriate information access and clear responsibility structures
5. Explainable and Interpretable: Systems offer understandable representations of their operation and outputs
6. Privacy-Enhanced: Systems safeguard human autonomy, identity, and dignity
7. Fair with Harmful Bias Managed: Systems address equality and equity concerns while mitigating harmful bias

Trustworthiness is a social concept that ranges across a spectrum and is only as strong as its weakest characteristic. Creating trustworthy AI requires managing tradeoffs between characteristics - for example, between interpretability and privacy, or between accuracy and fairness.

The assessment of trustworthiness depends on context and the specific role of AI actors in the AI lifecycle. What constitutes trustworthy AI may vary between designers, developers, deployers, and end users of the same system.