Zenity Labs Uncovers Zero-Click Exploits in Major Agentic Browsers
Event summary
- Zenity Labs revealed zero-click PleaseFix exploit chains at Black Hat USA 2026, affecting Claude in Chrome, Gemini in Chrome, Perplexity Comet, ChatGPT Atlas, and Copilot Edge.
- Exploits range from data theft to account takeover and full machine control via 'Intent Collision' attacks.
- Research demonstrated bypasses of patches in Perplexity Comet and exploitation of agentic browser trust models.
- Some vendors patched vulnerabilities while others deemed findings as intended functionality.
The big picture
Zenity Labs' findings highlight a critical tension between convenience and security in next-generation browsers. The research exposes inherent vulnerabilities in agentic browser architectures that could reshape how enterprises approach AI integration. With major tech players involved, the industry faces pressure to standardize security protocols or risk widespread exploitation of these powerful but potentially dangerous tools.
What we're watching
- Security Model Shift
- How agentic browser architectures will adapt to fundamental security model changes.
- Industry Response
- Whether vendors will uniformly address PleaseFix vulnerabilities or maintain current functionality.
- Regulatory Scrutiny
- The pace at which regulators may intervene in AI agent security standards.
Related topics