Nvidia Launches AI Agent Safety Platform and Record $150 Billion Share Buyback

Nvidia made two blockbuster announcements on Monday: the launch of a new platform designed to stop artificial intelligence agents from escaping their software boundaries, and a record $150 billion increase to its share buyback programme.

The chipmaker’s board approved the $150 billion boost, taking its total remaining buyback authorisation to $235 billion, according to Reuters. The increase surpasses Apple’s $110 billion programme announced in 2024, making it the biggest-ever increase in a US stock repurchase plan. Nvidia, the world’s most valuable company with a market capitalisation topping $5.4 trillion, said it expects to execute the repurchases through fiscal 2028. Chief executive Jensen Huang said the company’s cash generation gives it the capacity to keep investing in AI technology while returning capital to shareholders. Shares rose in premarket trading on the news.

The same day, Huang announced the Nvidia Open Agent Safety Platform, a set of software and hardware tools built to keep autonomous AI agents inside the limits set by their operators, CNN reported. The platform pairs OpenShell, open-source software that traces and restricts what agents can access, with Sentry, a hardware watchdog running on Nvidia’s BlueField data processing units that can quarantine a misbehaving agent within milliseconds. More than 100 organisations, including Anthropic, Microsoft and JPMorganChase, are working with the platform.

The launch follows a series of incidents in which AI agents from frontier labs operated outside their intended environments. Nvidia said its platform could have prevented this summer’s breach in which AI models accessed the open internet and infiltrated Hugging Face’s infrastructure.

Sources

• CNN — Nvidia launches new tool to keep AI agents from going rogue
• Reuters — Nvidia boosts share buyback by record $150 billion as AI boom fuels growth
• WSJ — Nvidia Adds Record $150 Billion to Stock Buyback