AI Incident Database #1629 · 2026-07-30
Anthropic Research Model Reportedly Scanned 9,000 Targets and Compromised Real Company's Application During Evaluation
During an Anthropic cybersecurity evaluation with Irregular, an internal research Claude model reportedly scanned roughly 9,000 internet targets after failing to reach its fictional target.
- Risk class (MIT)
- Not yet classified
- Reports
- 16
- Developers and deployers named
- 6
On ShieldMap: Anthropic (named by AIID as a developer or deployer).
Reports
- An alignment assessment of recent cybersecurity incidents · anthropic.com, 2026-09-09
- AI models are breaking out of their cages. Their creators are scrambling. · washingtonpost.com, 2026-08-10
- What the latest rogue AI incidents should teach us · transformernews.ai, 2026-08-05
- Anthropic Finds Claude Breached Real Companies During Security Evaluations · securityaffairs.com, 2026-07-31
- Anthropic says its AI models hacked 3 organizations during testing · apnews.com, 2026-07-31
Title and description from the AI Incident Database (incidentdatabase.ai), snapshot of 2026-11-16, shared under CC BY-SA 4.0. Selected incidents; descriptions shortened; harmed parties and report text left out. The description is AIID's summary of what was reported, not ShieldPi's finding.