By Suhasini Srinivasaragavan
Publication Date: 2026-09-28 11:18:00
Jensen Huang wants the industry to self regulate on AI safety, says new laws would be unnecessary.
Nvidia has unveiled an open software platform that promises to better restrict AI agents from taking problematic actions in order to achieve its goals.
‘Open Agent Safety Platform’ – the latest from the Open Secure AI Alliance – aims to strengthen AI security from agent testing to deployment, with software that monitors and sets boundaries for agents running on CPUs.
One part of the stack, the OpenShell software, runs agents in a sandbox and turns operators’ instructions into a defined policy. The software is compatible with third-party compute platforms, including those from Arm and Intel.
Meanwhile, Sentry provides in-silicon security enforcement, giving it the ability to quarantine and stop misbehaving AI agents.
The launch is a direct response from Nvidia to a new kind of security threat where AI agents take concerning leaps in liberty to access data. Last week, Australia reported one such incident, where OpenAI’s agents hacked into a government site and even rewrote some internal files.
Later that week, the company confirmed similar incidents relating to a number of US government department websites. In all these cases, the agents were looking for information when they decided to circumvent security controls that stopped them from any further access.
“AI’s extraordinary potential for society will only be realised if…
