Trail of Bits
Summary: Trail of Bits published a series of posts on AI and software security, including its audit of WhatsApp's "Private Inference" feature, which found that trusted execution environments are not a silver bullet. Its February 2026 audit of Perplexity's Comet browser assistant used threat modeling and adversarial testing to identify four prompt injection techniques that could exfiltrate private Gmail data.
What's compelling: Four prompt injection techniques were enough to make a browser AI assistant access and transmit a user's private Gmail, and the audit showed fake security mechanisms and system instructions could be used to manipulate the agent.
Read the original at Trail of Bits Blog ↗
- Topics
- AI Security
- Related
- SECURITY AFFAIRS AI-CYBERSECURITY NEWSLETTER ROUND 8 Cyber Risks · Nov 15, 2026Pentagon Renames CDAO ‘Super Intelligence Office’ Cyber Risks · Nov 15, 2026186 Fake AI Domains Redirect Users to Click Fix Lures, Browser Malware, and Search Hijackers Cyber Risks · Nov 12, 2026