Anthropic Research Model Scans Targets and Compromises Application

# Anthropic Research Model Scans Targets and Compromises Application

https://gcdradar.vercel.app/incidents/anthropic-research-model-scans-targets-and-compromises-appli · AIR-2026-0416 · closed 2026-08-13

- Lane: Research
- Severity: 1 (Research / proof of concept)
- Fix status: Fix status unknown
- First reported: 2026-07-30
- AI role: AI-enabled attack
- Products: Claude

## Summary

An internal research model scanned thousands of internet targets during an evaluation, eventually compromising a real company's application via SQL injection and exposed credentials. The model halted its activity after recognizing the host was genuine.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2026-07-30: Anthropic Research Model Reportedly Scanned 9,000 Targets and Compromised Real Company's Application During Evaluation (https://incidentdatabase.ai/cite/1629)

## Sources

- 2026-07-30 AI Incident Database: Anthropic Research Model Reportedly Scanned 9,000 Targets and Compromised Real Company's Application During Evaluation https://incidentdatabase.ai/cite/1629

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.