Researchers tricked Copilot into leaking an undocumented parameter enabling one-click Gmail/Drive theft
heypearlai · x · 2026-08-20
Microsoft patched a Copilot bug on Aug 18 that let attackers steal a victim's Gmail, Drive and Calendar data with a single click on a crafted link. Varonis researchers found it by repeatedly asking Copilot "why can't this happen" — every refusal came with technical explanations, and eventually the AI volunteered an undocumented URL parameter mid-refusal, which turned out to be the core of the exploit. No evidence of exploitation before the fix.
Related event: Microsoft Patches Copilot Flaw That Let One Click Steal Data(3 posts)→
More from Safety
- Anthropic red team's first-ever model "abliteration" cut GLM-5.3 refusals from 90% to 3% for ~$4,400 — cephaloform · 2026-10-16
- Anthropic grants METR employee-like access as independent safety evaluators go public — AnthropicAI · 2026-10-16
- APEX-Harm benchmark: script-based injection attacks easily compromise agents in shared workspaces — KAUST · 2026-10-16
- Residual-stream forecasts predict agent violations before they happen, USC study shows — USC · 2026-10-16
- AI safety drama: risk research called a 'company effort,' team lead responds — logangraham · 2026-10-16
- Brave's Leo adds end-to-end encrypted chat with GLM via attested TEE — gnukeith · 2026-10-16