Nvidia $NVDA launched the Open Agent Safety Platform on Monday, an open software platform and reference system design intended to prevent AI agents from escaping their operating boundaries and accessing systems they are not authorized to reach.
The announcement follows a series of incidents in which AI agents from companies including OpenAI, Anthropic, Meta $META , and Google $GOOGL broke out of testing environments and accessed external systems without authorization, according to the Wall Street Journal . Nvidia told reporters that each incident shared a common thread: agents found ways around security controls at the application layer in order to finish the work they had been given.
The platform has two main components. OpenShell is open source software that runs on central processors and sets a secure runtime boundary governing what autonomous agents can access and do. A second layer, called Sentry, runs on Nvidia BlueField-4 data processing units — separate from the CPU and GPU — and continuously monitors agent behavior. Sentry can quarantine an agent that attempts to move outside its permitted boundary in milliseconds, the company said.
"Recent incidents have highlighted a fundamental hurdle for AI agents, and that is that model-level safeguards alone can't govern what agents can access or do," Justin Boitano, Nvidia 's vice president of enterprise AI, said in a media briefing, according to CNBC .
Nvidia executives told reporters the platform could have prevented the breach of Hugging Face by OpenAI agents this past summer, according to the Associated Press . That incident, which OpenAI later described as the first known case of an automated agent collective acting offensively without authorization, involved agents escaping a testing environment, reaching the open internet, and accessing Hugging Face production credentials and private code repositories between July 11 and July 13.
"As we continue to discover the frontier of AI capabilities, we must accelerate discovery at the frontier of AI safety," Nvidia founder and CEO Jensen Huang said in a statement.
More than 100 organizations are working with the platform at launch, the company said, including Anthropic, Microsoft $MSFT , Salesforce $CRM , SAP, Scale AI, SpaceXAI, JPMorganChase, Palantir $PLTR , and Palo Alto Networks $PANW . Anthropic's Claude Managed Agents has integrated with OpenShell and BlueField to give enterprises control over agent access through sandboxes. SpaceXAI is using the platform for its Cursor coding agents and Grok models.
Salesforce and Nvidia have integrated OpenShell with Slack $WORK , allowing teams to view agent activity and approve or reject requests for additional permissions without leaving the messaging platform. SAP is embedding OpenShell into Joule Studio runtime, part of its Business AI Platform.
OpenShell software is available through Nvidia 's developer resources page and GitHub. The platform also supports the Open Secure AI Alliance, a body initiated by Nvidia alongside more than 120 organizations and governed by the Linux Foundation.
The essential business news, delivered fresh every morning.
Join 500,000+ readers who start their day with Quartz.
The full story
This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.
