Trustworthy AI systems are artificial-intelligence systems designed and operated to be reliable, safe, transparent, fair, accountable, and respectful of privacy throughout their lifecycle. The concept, codified in frameworks such as the NIST AI Risk Management Framework and the EU’s trustworthy-AI guidelines, integrates technical robustness with governance so that stakeholders can justifiably rely on the system’s behaviour.
Content
- Trustworthiness is multidimensional, spanning robustness to distribution shift and adversarial attack, explainability, bias mitigation, privacy preservation, and clear lines of human oversight. Standards bodies operationalise these properties into assessment and risk-management processes so that deployment decisions can be evidenced rather than assumed.