Score an AI red-team finding using impact, likelihood, exposure, and compensating controls.
A customer-success copilot drafts an incorrect refund policy once in 25 red-team attempts. Impact x likelihood x exposure, adjusted by control strength The common trap is severity by emotion: finance words sound scary, so the finding becomes critical without checking frequency, reach, or approvals. Impact If used, the bad draft could promise a refund above policy and create customer or revenue impact. Impact is real because the content could change a customer commitment. Likelihood Observed 1 failure in 25 seeded attempts, tied to ambiguous refund-threshold prompts. The denominator keeps the finding honest while preserving the failure condition. Exposure The copilot…
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in