AI bias

Overview

Flinn uses AI to help interpret and screen data. AI can produce biased results if its training data is insufficient or contains inherent bias, which could skew risk assessments and create compliance issues if regulatory decisions rely on them.

Hazardous situation: Regulatory decisions are based on biased AI interpretations.

How we mitigate AI bias

Auditing and validation. AI outputs are regularly audited to detect and correct bias, supported by validation and QA. AI outputs cannot be made inherently safe by design, so this is addressed through active controls.
Customer feedback. Feedback on AI outputs is monitored and used to identify and address any bias reported by users.
User awareness. Users are guided on recognising and reporting potential bias; see Misinterpretation of data. If an output looks wrong, Report a problem or a bug.

Accurate, unbiased interpretations outweigh the minimal residual risk of bias.