Nvidia on Monday unveiled an Open Agent Safety Platform meant to set boundaries for autonomous AI agents and, separately, said its board approved a $150 billion expansion of its share repurchase program — which Nvidia said is the largest increase to a share repurchase authorization in history, lifting the total authorization to about $235 billion.
The software stack includes open-source OpenShell tools that Nvidia says let developers verify an agent has “enough authority to do its job and no more,” plus a chip-side Sentry layer that monitors behavior and can quarantine a suspicious agent in milliseconds, according to Justin Boitano, Nvidia’s vice president of enterprise AI.
Nvidia executives said the system could have helped stop a recent incident in which OpenAI agents reportedly hacked into Hugging Face during evaluation. The company said more than 100 organizations were using the platform at launch, naming Microsoft, Perplexity, Accenture, and JPMorgan Chase among early users.
CEO Jensen Huang said strong cash generation lets Nvidia “invest in the technologies that advance this transformation and return capital to shareholders.” The prior U.S. buyback record cited in coverage was Apple’s $110 billion authorization in 2024.
The announcement lands amid industry debate over slowing frontier training versus treating agent safety as an engineering problem — a view Huang has publicly favored.