Menlo Security expanded its Menlo Agent Runtime Security (MARS) platform to secure AI assistants, coding agents, and autonomous agents, including Microsoft Copilot, Google Gemini, Claude Code, and Claude Cowork. MARS, a cloud-based Browser Security Platform introduced earlier this year, protects agents by sanitizing the web pages and files they read, neutralizing prompt injection, and blocking data exfiltration. It is the first platform to strip prompt injection and hidden instructions from agent-read content and prevent sensitive data leaks. The release addresses threats like instructions hidden in white-on-white text, file metadata, or a single human-invisible pixel. Mandiant’s 2026 AI Risk and Resilience report names prompt injection a top threat to AI applications. Gartner projects that by the end of 2026, 40% of enterprise applications will include task-specific AI agents, up from less than 5% in 2025. MARS runs agent web activity in the Menlo Cloud, cleaning pages and files, including downloads and uploads, before the agent sees them. Bill Robbins, Chief Executive Officer of Menlo Security, stated that agents operate at machine speed without human skepticism and that MARS provides runtime guardrails so security teams can approve agentic AI.