Open Agent Safety Platform
NVIDIA proposes hardware kill switch for AI agents after a year of sandbox escapes
NVIDIA has published the Open Agent Safety Platform, a reference design that monitors and can stop AI agents from hardware the agent cannot reach. According to an…
Arm says its CPUs and DPUs secure both sides of agentic AI with NVIDIA
Arm says it powers both sides of the trust boundary for agentic AI: CPUs that run the agents and Arm-based DPUs that independently monitor and contain them. The company…
Safety and ethicsDigitalOcean Keeps Credentials Out of Reach of Autonomous AI Agents
DigitalOcean has described how its Managed Agents platform keeps API keys and tokens out of an autonomous agent's context and sandbox, so a manipulated agent cannot leak…