Gen launches Agent Trust Hub to gate AI agent actions
Original title: You gave your AI agent access, but who is giving it orders?
The hub adds skill scanning, an agent VPN, Sage runtime hooks, and Norton monitoring between an agent’s decision and execution.
Gen describes Agent Trust Hub as a trust layer for AI agents, covering verification, connectivity, runtime checks, and data visibility. Gen Threat Labs says a controlled test steered an agent via email content into running commands, sending data out, and deleting files, without a software vulnerability. The company also cites Warden Stealer, malware aimed at AI assistant and coding-tool configs, tokens, and chat histories. Sage, which is open source, can allow, prompt, or deny shell, file, and web actions on supported agents.
Unconfirmed: The 12% malicious-skill figure and the agent experiment are Gen Threat Labs’ own findings.
- Gen researchers said 12% of ClawHub skills were malicious at the time checked.
- Sage covers Claude Code, Cursor, VS Code, OpenClaw, and OpenCode.
- VPN for Agents uses Norton VPN infrastructure and separate tunnels per agent.
- Gen published AARTS, an open spec for agent runtime safety events.
- Norton AI Agent Protection in Norton 360 monitors supported agent activity.
Impact: Agent actions and exposed assistant credentials, tokens, and chats
- 1.Review agent skills and plugins before install using available scanning or risk ratings.
- 2.Intercept agent shell, file, and web actions before execution where the platform allows it.
- Vendors
- Gen
- Malware
- Warden Stealer
- Organizations
- Gen Threat LabsVercel
From the original article
5 min read at bleepingcomputer.com
AI agents can be manipulated through malicious emails, websites, and plugins into executing commands, stealing data, or deleting files using their legitimate permissions. Gen Digital explains how its Agent Trust Hub helps protect agents through skill verification, secure connectivity, and runtime controls that intercept dangerous act [...]
By Sponsored by Gen Digital · bleepingcomputer.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.