00:00NVIDIA is rolling out a new software platform to help keep AI agents within their limits
00:06and stop them from accessing systems they are not supposed to.
00:10The move comes as concerns grow over how much freedom AI agents should have
00:15following recent incidents involving models from OpenAI and Anthropic,
00:21raising concerns about the risks of more autonomous AI systems.
00:26NVIDIA's open agent safety platform includes tools that restrict what AI agents can do
00:32and monitor their activity.
00:34The company said it is launching the tools with partners including Dell, Microsoft, Oracle and Anthropic.
00:41The move comes amid multiple reports of AI agents going rogue.
00:45Most recently, Australia's government said an OpenAI agent had accessed a government health portal
00:51in what it described as the first such incident.
00:55Amid calls for greater AI safety, NVIDIA CEO Jensen Huang has rejected broad safety regulation,
01:01saying the risks can instead be addressed through engineering and safer system design.
