Nvidia's board investigates its CEO's contacts as House chairs summon him

The Nvidia board said on 8 November that it had begun an independent investigation. The inquiry follows reporting that founder and chief executive Jensen Huang was the target of a foreign intelligence effort. A committee of independent directors, assisted by outside counsel, will oversee the inquiry. The board said it had no reason to believe that national security or Nvidia secrets were compromised, and that Huang supports the review and has committed to cooperating.

The reporting centers on Huang’s contacts with Alice Huang, an executive at Megaspeed, a Singapore-based cloud company that has bought billions of dollars of Nvidia chips. The New York Times reported last week that US intelligence officials believed Huang had been targeted by a Chinese espionage operation, and that their concerns centered on that relationship. The paper said Huang had not received the defensive briefing usually offered to potential targets. Nvidia rejected the account. “Meetings with customers are part of a CEO’s job,” the company said, adding that regional cloud customers receive ordinary business information, not trade secrets. The chairs of the House committees on China, foreign affairs and intelligence have asked Huang to appear at a closed joint briefing on 18 November. Nvidia would not say whether he will attend.

Featured source: Ars Technica , Nvidia board statement .

Other Stories

Chainalysis says an intruder took credentials and tokens for every customer

The blockchain analysis firm told customers that an October breach exposed names, usernames, password hashes and live tokens for all users of its platform. Chainalysis said the attackers used AI agent-assisted offensive techniques, and that one part of the campaign was an on-path attack that exposed a subset of customer passwords. The company said it had identified and fixed that vector. Its customers include the FBI, Immigration and Customs Enforcement, the US Air Force, Microsoft, Coinbase and Barclays, according to Chainalysis and public records. Access to its know-your-transaction and address screening services is returning gradually, with mandatory credential resets and allow-listed IP addresses. The company has not said whether customer investigation data was taken.

Filed from: Ars Technica .

Four AI companies will treat one robots.txt line as a training opt-out

Google, Microsoft, OpenAI and Anthropic say they will treat a shared policy token in robots.txt as a site-wide preference about AI training. The token replaces the per-crawler rules publishers write today. It comes from a draft in the IETF’s AI preferences working group, which each company has described implementing on its own timetable. Anthropic has started, OpenAI has promised to follow in the coming months, Microsoft starts on 1 February 2027 and Google is targeting 2027. The draft is explicit that the token is not a directive about crawling. Bots may still fetch pages, and a site that says nothing continues to allow training. Meta’s communications chief called the signal an important step, but the company has not committed to honoring it. Google plans to read an opt-out as also disallowing its Gemini training and grounding crawler. That applies unless a site names Google-Extended separately, and it can remove a publisher from AI answers.

Filed from: TechRadar , Google Search Central , Microsoft Bing .

A database of 10,202 local comments finds 90.5% opposed to data centers

American Sentiment, a site run by the San Francisco startup Prophecy Gov, collected public comments from 906 meetings in 417 local governments across 48 states. It found 90.5% opposed to data center projects, 7% in support and 2.5% asking questions, with no supportive comment recorded in 73% of the meetings. Water use drew the most objections, followed by electricity costs and pollution. Only 21% of submissions mentioned artificial intelligence, so most objections concerned land, water, power and process, not the technology itself. About 79% of commenters who identified as union workers supported construction, compared with 3% of property owners and residents.

Filed from: TechRadar .

RTX 5090 cards sell far above list price as retail stock empties

Nvidia launched the GeForce RTX 5090 at $1,999. Newegg now lists MSI's Suprim card at $5,899.99 and an Asus TUF model at $5,099.99, both unavailable, according to TechRadar. B&H Photo Video shows two 5090 cards out of stock at $4,599.99 and $4,499.99. Amazon offers an Asus TUF card at $7,679.80, about 284% above the launch price. Lenovo and HP no longer list readily purchasable desktops with the card. Prebuilt machines from smaller manufacturers remain available, from about $9,000 to more than $13,000. Nvidia has not confirmed that it has ended production, and the 32GB card is also used for local AI work.

Filed from: TechRadar .

Envision starts up an 800-volt DC distribution system at a Chinese AI campus

Envision says an integrated 800-volt direct-current power architecture is now in commercial operation at its Galaxy AI campus in Ulanqab, Inner Mongolia. Trials of its Dubhe controller have run there since August. A solid-state transformer converts 10-kilovolt alternating current from the grid into an 800-volt DC line, skipping conversion stages that conventional halls use. Battery storage responds to sudden swings in accelerator demand and provides backup measured in minutes. The company says the design cuts losses, uses less copper and fits more racks into a given floor area. No independent measurements of efficiency or cost have been published. Envision expects the campus to reach 3 gigawatts.

Filed from: TechRadar .

Instagram widens AI age checks to every market

Meta is expanding the age estimation system it uses to find teenagers who gave Instagram a false birth date. The rollout follows launches in the United States, the European Union and Brazil. The tools weigh signals such as the accounts and posts a user interacts with, birthday messages received and visual cues in photos and videos. Those cues include height and bone structure. Users judged to be 13 to 17 are moved into Teen Accounts, which are private by default and block messages from strangers. Meta removes users it believes are under 13. In Australia it now also uses Apple’s age-range signal, and parents in the United States will receive an explanatory notice next month.

Filed from: The Verge .