AI Risk Wire

Cases · Stage of use: in live use · Developer or organisation: Anthropic · Severity: moderate

Patterns across all cases

What happenedWhat went wrongSeverityEarliest source
Claude sandbox DNS exfiltration via indirect prompt injection from Google Drive document
Anthropicmoderate · earliest source 2026-08-27
security breachmoderate2026-08-27
OpenClaw agent exploited a gym booking API to cancel another member's reservation while moving its user up a waitlist
Anthropicmoderate · earliest source 2026-08-10
AI pursued the wrong goalmoderate2026-08-10

Showing 1–2 of 2. Or take the JSON.