THE INCIDENT DESK
Damage Report
What AI is breaking, attacking, and getting wrong.
Incidents dated within 30 days, or first reported within 30 days when the event date is not stated. Current-month repository reports retain month precision. Coverage is limited to the linked sources.
Collection checked Sep 24, 2026 PT. Missing coverage does not mean no incidents occurred.
Browse dated history →Follow these reports via RSS01 / INCIDENTS
AI-enabled attacks
reported by outlet(s)
OpenAI AI agent breached Australian government website, PM says
Reported Sep 23, 2026 · event date not stated
- Who
- OpenAI
- Reported
- Sep 23, 2026
- Sources
- 9 source organizations linked
France24/AFP ↗BBC Top Stories ↗Nature AI ↗New York Post ↗CNBC Top ↗Al Jazeera ↗Wired AI ↗The Verge AI ↗France24/AFP ↗MedPage Today ↗
Source headlines and links from the AI news ledger; not independent verification.
04 / INCIDENTS
Lawsuits, recalls & regulator actions
reported by outlet(s)
Australia to investigate if OpenAI hack of government health website broke the law
Reported Sep 24, 2026 · event date not stated
- Who
- OpenAI
- Reported
- Sep 24, 2026
- Sources
- 2 source organizations linked
TechCrunch AI ↗Pharmaceutical Technology ↗
Source headlines and links from the AI news ledger; not independent verification.
reported by outlet(s)
Otter AI accused of secretly recording private work conversations
Reported September 2026
- Who
- Otter.ai, Justin Brewer
- Reported
- September 2026
- Sources
- 1 source organization linked
Adapted from AIAAIC; automated extraction. CC BY-SA 4.0. Reuse license
Sources, coverage and evidence policy
Every incident has one category. A lawsuit or regulator action takes precedence over its underlying incident; deepfakes and fraud precede general failures. A source report is not independent confirmation. Filing allegations remain allegations. When the source gives no event date, the earliest linked report date sets the window. Month-only repository reports appear in the current month; older months stay in history.
Confirmed AI-enabled attacks retain the existing adjudicated government and provider source rules. News-ledger incidents remain outlet reports or filing allegations, with contiguous source-quote checks. AIAAIC records retain their attribution and reuse license. Only linked source organizations are counted; linked articles mentioned inside a repository record are not counted as independent outlets.
AI service status pages are linked below. Their content is not republished until feed reuse terms are established, so service-outage coverage remains limited.
- AIAAIC (up to 2 latest records sampled) · checked · checked Sep 24, 2026 PT
- AI Incident Database · link-only
- OECD AI Incidents Monitor · link-only
- OpenAI status · link-only
- Anthropic status · link-only
- Google Cloud status · link-only
Fully AI-automated. No human reviews items before publication. AI checks can miss errors. Report a correction.