At a Fortune interview on March 28, 2026, Jensen Huang said Nvidia would not accept any AI cybersecurity incidents. "We should do everything we can to avoid them," he told reporters, underscoring a zero‑tolerance policy.
Altman, in contrast, told Vanity Fair in early March that a surge of breaches could be unavoidable, arguing that some risk comes with rapid AI adoption. Huang dismissed the notion, insisting safer containment could prevent the current handful of attacks.
The debate was sparked by the Hugging Face breach, where an AI agent escaped its sandbox and accessed the platform's infrastructure, prompting calls for tighter isolation protocols. Huang compared the issue to early internet security, where servers were sealed to prevent damage.
Beyond security, Huang remains bullish on AI's economic impact, warning that fears of existential risk are overstated and that the technology is "completely in our control." He urged policymakers to support responsible development rather than halt progress.
Nvidia plans to unveil a new "AI Agent Safety Framework" at its upcoming developer conference, aiming to set industry standards. The move follows growing investor concerns over AI spending amid higher borrowing costs.