Sunday’s AI news arrived with an unusually clear theme: trust is now an operating requirement, not a line in the launch deck. In AI news today, Nvidia’s board opened an independent investigation after reports that CEO Jensen Huang was targeted by a foreign intelligence effort. The board says it currently has no reason to believe company secrets or U.S. national security were compromised, but it is bringing in outside counsel anyway. Elsewhere, Anthropic reportedly moved to secure 1.7 gigawatts of gas-fired generation, Australia’s central bank found little aggregate productivity or employment impact from AI so far, and questions grew around the independence of the U.S. government’s AI-safety center. Add fresh litigation over AI-generated music and Amazon’s reported return to the home-robot market, and the day’s lesson is hard to miss: capability is moving quickly, while institutions are still building the controls, evidence, and accountability needed to live with it.

AI news today: Nvidia’s board opens an independent security investigation

Nvidia’s board said on November 8 that an independent-director committee and outside counsel will examine reporting that founder and CEO Jensen Huang was targeted by a foreign intelligence effort. The company emphasized that it has no present reason to believe Nvidia secrets or national security were compromised, and said Huang supports the review.

That is the confirmed part. CBS News reported that the underlying allegations concern a Chinese business executive whom U.S. intelligence officials reportedly suspected of working for Beijing; China has rejected the espionage allegations. For customers and investors, the immediate question is not personal gossip but governance at a supplier central to restricted AI hardware. Watch the inquiry’s scope, any findings about access to sensitive information, and whether export-control compliance changes.

Anthropic reportedly secures 1.7 gigawatts of gas-fired generation

The Financial Times reported on November 8 that Anthropic agreed in late September to pay about $3.4 billion for commitments covering 1.7 gigawatts of Caterpillar gas-fired generators originally ordered by OnePWR Solutions. The reported transaction includes engineering and construction capabilities, but not OnePWR itself or its employees. Anthropic and OnePWR declined to comment, so this remains a reported deal rather than a company-confirmed announcement.

The move would give Anthropic more direct control over a scarce input: dependable power. It also shifts part of the AI buildout from cloud contracts toward energy development, with emissions, permitting, and financing attached. The scale fits the compute obligations examined in our November 3 brief. What comes next is evidence: project sites, operating timelines, fuel arrangements, and whether the generators sit behind the meter or support the wider grid.

Australia’s central bank sees adoption—but little aggregate AI payoff yet

In a speech delivered Monday morning in Sydney—Sunday evening in the United States—Reserve Bank of Australia Assistant Governor Sarah Hunter said AI adoption is spreading quickly, but there is limited evidence that it has materially lifted aggregate productivity. Estimates for AI’s contribution to annual productivity growth over the next decade range from roughly 0.1 percentage point to more than 2 percentage points.

The RBA’s business liaison found around three-quarters of firms reported no AI-related headcount change, while smaller shares reported slower hiring, reductions, or increases. The bank cautions that exposure to AI is not the same as job loss, and that tentative weakness among younger workers in highly exposed occupations may have other causes. For leaders, this is a useful antidote to both panic and triumphalism: measure task-level gains, reorganization costs, and hiring changes before declaring an economy-wide result.

U.S. AI-safety officials reportedly begin informal White House reporting

Associated Press reporting says staff at the Commerce Department’s Center for Advancing Innovation and Standards for Super Intelligence, or CAISSI, were asked to report informally to a White House office. The account relies on two former officials speaking anonymously. AP also reports that the $10 million-a-year center lacks a permanent leader and has stopped publishing its research; the White House did not answer specific questions about the arrangement.

Closer White House access could bring money and influence, or weaken the perceived independence of evaluations involving powerful labs. Anonymous-source reporting cannot resolve that tradeoff. Watch for a formal reporting structure, permanent director, publication policy, and technical results outsiders can examine.

Elvis and Eminem rights holders test AI music under publicity law

Rights holders connected to Elvis Presley and the publisher of several Eminem recordings have sued Suno, alleging that its system produces unauthorized imitations of the artists’ voices, music, and lyrics. Sunday’s reporting by The Verge highlights the Presley plaintiffs’ use of Tennessee’s ELVIS Act, which explicitly protects voice and likeness. A case summary of the November 6 complaint says it also alleges training on 859 compositions. Those are allegations, not findings.

Suno says it prohibits unauthorized use, blocks prompts naming specific artists or songs, and screens uploaded audio and lyrics. The cases matter because they test more than copyright: a model might create something technically new while still sounding recognizably like a protected person. Watch how courts separate voice, style, composition, training data, and generated output.

Amazon reportedly assembles a second home-robot push

Amazon has formed a Personal Robotics Group and is recruiting for what job listings call a “new-to-world robot product,” according to CNBC’s analysis of hiring posts. The effort follows the limited, invitation-only Astro robot and Amazon’s acquisition of Fauna Robotics, maker of the 3.5-foot Sprout humanoid. Amazon confirmed its commitment to personal robotics but has not disclosed the new product’s design, tasks, price, or release date.

The living room is not a factory floor. Pets, stairs, children, clutter, privacy, and support costs turn a polished demo into a reliability test. Amazon’s warehouse experience helps, but consumer value will depend on useful physical tasks. Treat this as a reported development program, not a launch.

Watch & Learn

Editor’s note: Leo Cui’s 41-minute explainer, The Entire AI Data Center Explained — From Electricity to ChatGPT, follows one prompt through power, cooling, GPUs, memory, networking, and software. It is best for operators, investors, and curious builders who want the whole infrastructure map rather than another chip benchmark. Cui is a venture investor and clearly discloses that perspective.

AI, Translated: behind-the-meter generation

Behind-the-meter generation means producing electricity on the customer’s side of the utility meter, often at or beside the facility using it. A data center might run on-site gas generators instead of drawing every megawatt from the public grid. That can shorten waits for grid connections and improve resilience, but it does not make emissions, fuel supply, permitting, or noise disappear. You should care because AI companies increasingly compete not only for chips, but for the power systems that keep those chips running.

Try This Today: make Claude audit a reported company plan

Goal: turn a reported plan into a decision-ready fact sheet without blurring claims and confirmation.

1. In Claude, select Research and make sure web search is enabled.
2. Paste the reported claim and ask for a source ladder: filings and company statements first, reputable reporting second.
3. Require separate sections for confirmed facts, attributed reporting, conflicts, and unanswered questions.

Copy-ready prompt: “Research this claim: [paste claim]. Build a dated source table. Prioritize primary records, then independent reporting. Label every conclusion Confirmed, Company claim, Reported, Disputed, or Inference. Flag circular sourcing and state what evidence would change the assessment. End with five questions for management.” Anthropic’s current documentation describes Research as a beta feature on paid Pro, Max, Team, and Enterprise plans; it requires web search and may use limits faster than an ordinary chat.

One thing to remember

The more consequential an AI system becomes, the less “trust us” can carry the load. Boards, labs, regulators, courts, and households all need the same thing: claims that can be checked against evidence.