Human Oversight Use Cases

AI diagnostic tools can surface patterns clinicians might miss, but they also hallucinate and lack the contextual judgment trained physicians apply.

Machine learning credit models can deny loans to creditworthy applicants based on proxies correlated with race, postcode, or gender without any explicitly discriminatory rule.

Large language models providing legal guidance confidently state incorrect case law, cite non-existent statutes, and misapply jurisdiction-specific rules.

GeraClone allows individuals to create AI replicas of themselves for productivity, expert monetisation, and legacy purposes.

Autonomous and semi-autonomous driving systems make split-second decisions that can be fatal.

AI financial advisers can recommend unsuitable products, misread risk appetite, and generate advice that exposes providers to regulatory liability.

Risk assessment algorithms used in bail, sentencing, and parole decisions have been shown to exhibit racial bias and produce wildly different scores for similarly situated defendants.

Automated benefits assessment systems — used for Universal Credit, disability assessments, and tax credit calculations — have generated mass errors that force vulnerable claimants into hardship.