Nvidia Launches Open Agent Safety Platform to Secure AI Agents
Nvidia has introduced its Open Agent Safety Platform (OASP), an initiative designed to secure autonomous AI agents from misbehavior and breaches. Announced through its newsroom and developer blogs, the platform features a two-layer security approach: OpenShell, an open-source software runtime, and Sentry, a hardware-level monitoring system integrated into BlueField-4 DPUs. This system aims to provide continuous in-silicon agent monitoring and runtime controls, ensuring AI agents operate within defined safety boundaries. Nvidia claims the OASP could have prevented recent high-profile incidents, such as the Hugging Face hack, by quarantining rogue agents in milliseconds. The company emphasized the importance of external enforcement mechanisms, rather than relying solely on model safeguards. The initiative is backed by over 100 industry partners, including companies like TrendAI and DigiCert, who are extending the platform with threat intelligence and identity controls, underscoring a broad industry focus on practical AI safety and security measures.
186 articles from 155 outlets covered this story. No difference in framing or figures was found between them. The underlying claim is sourced from a press release.
What do all outlets agree on?
155 outlets covered “Nvidia Launches Open Agent Safety Platform to Secure AI Agents”. All of them report the following:
- Nvidia launched a new AI safety platform
- The platform is called the Open Agent Safety Platform (OASP)
- It aims to secure autonomous AI agents
- It includes OpenShell (software) and Sentry (hardware)
- It is designed to prevent agents from going rogue or misbehaving
- It is an open-source initiative
- Over 100 industry partners are involved
- Nvidia claims it could have prevented past AI security incidents, such as the Hugging Face hack
- The system is designed to contain agents quickly, in "milliseconds"
Did outlets disagree about this?
No. All 155 outlets covering “Nvidia Launches Open Agent Safety Platform to Secure AI Agents” reported it the same way — no difference in framing or figures was found between them.
Which outlets covered this?
All 186 articles found on this story, grouped by the stance of the piece. Every link goes to the original publisher.
optimist8
pragmatist86
safety92