NVIDIA has launched the Open Agent Safety Platform, a new software and hardware reference system designed to give developers greater control over autonomous artificial intelligence agents, according to ChannelsTv and Reuters report. The platform comes as OpenAI, Anthropic, Meta and Google have disclosed recent incidents involving AI systems escaping controlled environments, attempting to access external systems or carrying out actions beyond their intended objectives. NVIDIA says its approach is designed to establish boundaries around AI agents from testing through deployment, as businesses increasingly use autonomous systems for software development, research, cybersecurity and other complex workflows.
At the centre of the platform are two technologies, NVIDIA OpenShell and NVIDIA Sentry. OpenShell is an open source runtime that allows developers to define and enforce policies governing what an AI agent can access and do, including restrictions around files, networks, applications and sensitive information. Sentry provides a separate monitoring layer running on network infrastructure, designed to continuously observe agent activity and intervene when behaviour moves outside established boundaries. NVIDIA said the two layers are intended to work independently, with OpenShell governing an agent’s authorised actions while Sentry provides an additional line of defence against suspicious behaviour. More than 100 companies are reportedly using the system at launch, including Microsoft, Perplexity, Accenture and JPMorgan Chase.
The launch highlights a growing commercial requirement for AI safety infrastructure as companies move from conversational AI towards agents capable of taking actions independently. NVIDIA said its security platform could have prevented a recent incident involving AI agents that breached the security of AI platform Hugging Face, although that remains the company’s assessment rather than an independently established outcome. The company is positioning Open Agent Safety Platform alongside its wider NVIDIA Agent Toolkit, which includes Nemotron models, OpenShell and other tools for building enterprise AI agents. With autonomous systems increasingly being deployed across finance, software, healthcare, cybersecurity and industrial operations, the ability to govern, monitor and contain AI agents is emerging as a significant part of the enterprise AI market.
