Claude for Chrome Extension Flaw Lets Rogue Add‑Ons Read Gmail, Docs & Calendar
What Happened — Researchers identified a logic flaw in the Claude for Chrome browser extension that allows any other extension with script‑execution rights on claude.ai to invoke Claude tasks targeting a user’s Gmail, Google Docs (including comments), and Calendar data. The issue stems from insufficient isolation between extensions, enabling a malicious add‑on to harvest or act on sensitive Google Workspace content.
Why It Matters for Compliance & Audit Readiness
- The scenario exemplifies a failure of SOC 2 Access Controls: unauthorized code can bypass intended user‑level boundaries and access protected data.
- Continuous‑compliance programs must evidence extension vetting, least‑privilege policies, and monitoring of third‑party code to demonstrate due diligence during audits.
- Verisq’s SOC 2 Access Controls capability provides automated evidence collection for extension‑management policies and real‑time alerts on anomalous script activity.
Who Is Affected — SaaS providers and enterprises that rely on browser‑based AI assistants, particularly those integrating Claude for Chrome within Google Workspace environments.
Recommended Actions
- Review and tighten your organization’s browser extension allow‑list; enforce “run‑only‑trusted‑extensions” policies.
- Map the flaw to SOC 2 CC6.1 (Logical Access) and CC6.2 (Least‑Privilege) controls, and collect evidence of policy enforcement.
- Deploy security‑awareness training that highlights risks of installing unvetted extensions.
Source: The Hacker News
Technical Notes
- Attack vector: malicious browser extension leveraging shared script context.
- Impact: potential unauthorized read of Gmail messages, Google Docs content, and Calendar events.
- Mitigation status: Anthropic restricted the arbitrary‑prompt path in May, but the extension‑level isolation issue remains.