Incident reporting captures AI failures including fabricated citations, incorrect extractions that reached work product, confidentiality exposures, and near misses caught before consequence. Reports feed remediation and pattern analysis.
It parallels incident response in other risk domains.
Alternative Names:
AI Incident Response, AI Failure Reporting
Why it Matters?
Without a reporting mechanism a firm learns about AI failures only when they cause visible harm, which forfeits the near misses that would have revealed the pattern earlier. A single fabricated citation caught in review is a data point; the same failure occurring across multiple matters is a system problem requiring different action. Reporting also produces the record demonstrating that the firm monitored its tools, which matters if supervision is later questioned.
Frequently Confused with
Related terms
Frequently asked questions
What should be reported?
Why capture near misses?
