As AI world debates security, NVIDIA releases open source tools for agents
Nvidia open-sourced the Open Agent Safety Platform, pairing OpenShell agent sandboxing with BlueField-4 DPU out-of-band monitoring; 100+ firms committed.
NVIDIA released the Open Agent Safety Platform, an open software platform and reference system design for governance and control of AI agents across software, hardware, compute, and robotics. OpenShell, built on Apache 2.0 open source, lets operators define the files, networks, tools, processes, and credentials an agent can access and test guardrails before enterprise deployment. BlueField-4 DPU security updates enable out-of-band monitoring of agent behavior and security policy enforcement. More than 100 organizations including Anthropic, Arm, Microsoft, Palantir, and JPMorgan Chase have committed to using it.
- OpenShell (Apache 2.0) restricts agent access to files, networks, tools, credentials
- BlueField-4 DPU updates add out-of-band monitoring and policy enforcement
- Over 100 organizations committed, including Anthropic, Microsoft, Palantir, JPMorgan
- Huang argues sandbox escapes are a technically solvable engineering problem
From the original article
5 min read at cyberscoop.com
One expert told CyberScoop that the announcement reflects industry recognition that after years of training models to behave safely or ethically, more outside controls are needed. The post As AI world debates security, NVIDIA releases open source tools for agents appeared first on CyberScoop.
By djohnson · cyberscoop.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.