Norton Adds AI Security Controls to Protect Files and Block Unsafe Actions
Norton, a provider of consumer cybersecurity and identity protection software, has introduced Safe AI, a collection of security features in Norton 360 designed to protect users interacting with AI assistants and agentic applications.
The offering includes Action Shield, which aims to block potentially harmful AI actions and prompt injection attempts. It currently supports Claude, Cursor, OpenClaw and Visual Studio Code.
Safe AI introduces three capabilities designed to help make using AI tools safer with protection against unwanted AI actions, it gives people more control over AI file access, and spots risky links before you click.
Safe Zone provides controls over which files AI applications can access, allowing users to assign full, read-only or blocked permissions. The feature is currently limited to Windows.
A third capability, Safe Links, identifies potentially dangerous URLs presented through browser-based versions of ChatGPT and Claude.
Norton cited research from parent company Gen showing a 54% increase in malicious content delivered to its users through AI chatbots over the preceding 90 days.
For enterprise security teams, the announcement highlights emerging endpoint risks as AI assistants gain access to local files, development environments and automated workflows.
Safe AI is available in Norton 360 for Windows and Mac, although individual feature support varies by platform and application.
Posted by Pure AI Editors on 10/26/2026