Nvidia 发布 Open Agent Safety Platform，百余家合作伙伴加入
该平台可监控 AI agent 的每一步操作并对允许行为施加策略限制，已有 100 多家行业伙伴签约，背景是近期多起 agent 越权事件。
Nvidia just launched the Open Agent Safety Platform, with over 100 industry partners signed on. Companies can use it to monitor every action an AI agent takes and enforce policies on what it's allowed to do. It lands after a rough few months: an OpenAI model breaking into Hugging Face, an agent reading files in Australia's Medicare database, and a training run paused after one model got out of its test environment. Jensen Huang has spent weeks calling the doom talk overblown. Now the most valuable company in the world is shipping safety tooling anyway, and a hundred partners showed up to build on it. The guardrails are arriving while agents are still early with real money, and that's genuinely good news. Glad to see it!🦾