Nvidia has stopped shipping powerful artificial intelligence chips to China, casting uncertainty over a Trump-backed export agreement as Washington and Beijing compete for AI dominance.
The U.S. chipmaker sold $2.4 billion worth of licensed Hopper-generation data center products to China last quarter but has no plans for additional shipments, interim CEO Colette Kress said during an earnings call Tuesday.
“We are no longer shipping licensed Hopper data center products to China,” Kress said, without explaining the reason for the halt or whether it would be permanent.
President Donald Trump eased restrictions on Nvidia’s H200 AI chips following a summit with Chinese President Xi Jinping last year, allowing exports to approved Chinese customers under strict licensing conditions.
But Beijing has also restricted imports of the advanced processors as it seeks to reduce reliance on American technology. Roughly 600,000 H200 chips were being held in Hong Kong earlier this month awaiting Chinese approval, Reuters previously reported.
Reuters contributed to this report.
Sen. Maggie Hassan, D-N.H., is vowing congressional hearings on artificial intelligence safety after a federal watchdog found that four major AI platforms generated detailed plans for terrorist attacks during government testing.
The Government Accountability Office (GAO) found that ChatGPT, Claude, Gemini and Grok each produced attack plans in less than 12 hours, including proposals involving explosives or attacks on the nation’s electrical grid.
Investigators were able to bypass the platforms’ safety safeguards, obtaining information about potential targets, timing and attack methods. Independent experts reviewed the plans and confirmed that some of the details could cause real-world harm.
“The AI public safety threat is no longer theoretical,” Hassan, the incoming chair of the Senate Homeland Security and Governmental Affairs Committee, said Wednesday. “If Congress does not act, it is a question of when — not if — a bad actor will use AI to plan and carry out a major attack.”
Hassan said one of the committee’s first hearings in the new Congress will examine AI-related national security risks, with lawmakers also considering legislation to strengthen safety testing before AI models are released.
The companies behind the four platforms acknowledged the importance of improving safeguards in responses to the GAO, though some noted that their systems have continued to evolve since the testing.
The Associated Press contributed to this report.
JPMorgan Chase is testing artificial intelligence agents capable of carrying out financial transactions without direct human intervention, potentially paving the way for autonomous AI systems to negotiate and exchange payments.
The banking giant and South Korea’s Hyundai Card completed an initial test in which AI agents executed a simulated foreign exchange transaction using JPMorgan’s blockchain-based payment infrastructure, CNBC reported.
The companies plan to expand testing in the first half of 2027, exploring whether multiple AI agents can negotiate, reach agreements and make payments directly with one another.
But while the technology has proven relatively straightforward to develop, JPMorgan officials say establishing safeguards and accountability for AI systems handling money independently remains a bigger challenge.
“Technology-wise, it’s really not the biggest hurdle,” Wee Kee Toh, global head of Kinexys Insights at JPMorgan, told CNBC of developing the technology itself.
To address those concerns, the bank has developed controls that restrict AI agents to predefined transaction limits, automatically blocking requests that exceed authorized amounts.
JPMorgan has not announced a timeline for making the technology commercially available.
{{#rendered}} {{/rendered}}The Trump administration is considering requiring U.S. companies to publicly disclose their use of Chinese artificial intelligence models as concerns grow over American reliance on Beijing’s technology, according to an exclusive Reuters report.
The Treasury Department and White House Office of Science and Technology Policy have discussed a proposal that would require companies using Chinese AI tools to disclose that information on their websites, three people familiar with the matter told the outlet.
The measure could affect companies like Airbnb, which uses Alibaba’s Qwen AI model to power its customer service chatbot.
The proposal aims to discourage reliance on Chinese AI without imposing outright restrictions on American businesses. It remains unclear whether the plan has gained support within the administration or will advance to formal rulemaking.
The discussions come as Chinese AI models, including DeepSeek, Qwen and Moonshot AI’s Kimi, gain traction among U.S. businesses as lower-cost alternatives to American-developed technology.
The issue has also drawn congressional scrutiny, with Republican lawmakers launching an investigation into U.S. companies’ use of Chinese AI amid concerns about national security and growing technological dependence on Beijing.
A cybersecurity startup used an advanced artificial intelligence model to uncover a critical security flaw in Java, a widely used software platform that supports systems at roughly 90% of Fortune 500 companies, according to The New York Times.
The startup, Calif, discovered the vulnerability using OpenAI’s GPT-6 Cyber model, revealing a weakness that could allow hackers to gain unauthorized access to sensitive computer systems.
The flaw, dubbed “Java4Shell,” affects technology used by banks, hospitals, government agencies and major corporations. Researchers demonstrated that it could be exploited to break into Keycloak, a widely used authentication system that protects organizational logins.
The Cybersecurity and Infrastructure Security Agency (CISA) urged organizations to “immediately patch” the vulnerability, while Oracle released a software fix Tuesday, the Times reported.
Calif CEO Thai Duong warned that unpatched systems could remain vulnerable for years, underscoring growing concerns about AI’s ability to uncover serious cybersecurity weaknesses faster than organizations can address them.
Sen. Mark Warner, D-Va., warned Tuesday that artificial intelligence may be contributing to a wave of major technology outages across the country, while emphasizing there is no evidence of a coordinated cyberattack.
Warner, the top Democrat on the Senate Intelligence Committee, said AI is allowing cybersecurity teams to identify vulnerabilities at unprecedented speeds, prompting companies and government agencies to rush security updates that can strain critical systems.
“What appears to be happening instead is a consequence of how quickly artificial intelligence is changing the cybersecurity landscape,” Warner said in a statement.
The warning follows widespread disruptions affecting Microsoft, major banks, airlines and the National Security Agency in recent days.
Warner called the outages a “wake-up call” for stronger AI safeguards and cybersecurity protections, criticizing President Donald Trump’s administration over cuts to the Cybersecurity and Infrastructure Security Agency (CISA).
Secretary of State Marco Rubio separately said Tuesday there was “zero evidence” linking the outages to hackers, foreign governments or other malicious actors.
{{#rendered}} {{/rendered}}Live Coverage begins here