Gen and Vercel Partner on AI Safety
Gen and Vercel have partnered to bring independent safety verification to the AI skills ecosystem. Gen's Agent Trust Hub will integrate with Vercel's skills.sh platform to provide risk verification for AI skills. The collaboration aims to protect developers and users from unsafe AI agents as they evolve into autonomous actors capable of executing real-world tasks.
- Gen's Agent Trust Hub will provide a four-tier risk classification for AI skills: Safe, Low Risk, High Risk, and Critical Risk. This is intended to give developers transparent risk ratings before they install or execute a new AI capability. - The collaboration is a response to security vulnerabilities in the growing AI agent ecosystem. Gen's Threat Labs identified over 18,000 instances of the open-source agent platform OpenClaw that were exposed to the internet and found that nearly 15% of the skills they observed contained malicious instructions. - Vercel's skills.sh is an open-source tool that allows AI agents to execute reusable actions, called skills, via the command line. The goal is to create a more controlled and auditable environment by separating the AI's reasoning from the execution of specific, predefined commands. - The integration will embed Gen's security verification directly into the skills.sh directory, which serves a community of over 6 million developers on Vercel's platform. - Gen, the company behind the Agent Trust Hub, is a global corporation that owns a portfolio of consumer brands focused on "Digital Freedom," including Norton, Avast, and LifeLock. - Autonomous AI agents introduce new security risks not as prevalent in traditional software, such as agent hijacking, cascading failures across connected systems, and the potential for agents to autonomously discover and exploit new vulnerabilities. They can also create privacy risks by accessing large amounts of personal or organizational data to function.
