Florida AG sues OpenAI, Altman | Warren opens Nvidia chip probe | Anthropic confidential S-1

Bottom Line Up Front

Florida Attorney General James Uthmeier sued OpenAI and Sam Altman in state court Monday, alleging the company concealed ChatGPT safety risks and citing two shootings whose suspects consulted the chatbot. Sens. Elizabeth Warren and Andy Kim called on Commerce Secretary Howard Lutnick to testify on Bureau of Industry and Security guidance covering chip exports to overseas affiliates of Chinese firms. Warren separately pressed Nvidia executives on export compliance and board oversight, citing Justice Department cases tied to $670 million in diverted chips and servers, with answers due June 18. Anthropic confidentially submitted a draft S-1 to the Securities and Exchange Commission, positioning the Claude developer ahead of OpenAI on a potential public listing after a $65 billion raise at a $965 billion valuation.

I.Top Stories

Florida AG sues OpenAI and Altman in state court, seeking personal liability over ChatGPT harms

Florida Attorney General James Uthmeier sued OpenAI and CEO Sam Altman in Florida circuit court Monday, accusing the company of concealing ChatGPT safety risks and aggressively marketing the chatbot while disregarding internal and external safety warnings, the New York Times reported. The complaint cites two separate shootings whose alleged gunmen consulted ChatGPT, including the 2025 Florida State University attack into which Uthmeier opened a criminal investigation in April. The state's complaint also references a Stanford Medicine study by psychiatrist Nina Vasan and the case of Adam Raine, a 16-year-old who his family says used ChatGPT to plan a suicide, per the Associated Press. OpenAI said in a statement its models repeatedly encouraged the individuals to seek real-world support and that it has cooperated with law enforcement in both cases. The complaint further alleges ChatGPT collects data from minors without meaningful parental oversight and causes behavioral addiction and cognitive harm.

Read at NYT ↗ • Read at AP ↗

Sens. Warren and Kim demand Lutnick testify on Commerce chip loophole guidance

Sens. Elizabeth Warren (D-Mass.) and Andy Kim (D-N.J.) on Monday called on Commerce Secretary Howard Lutnick to testify before the Senate Banking Committee about the Bureau of Industry and Security's handling of advanced AI chip exports, per Reuters. The senators issued the demand one day after Commerce posted guidance Sunday clarifying that license requirements apply to chip shipments to overseas subsidiaries of Chinese firms, including for Nvidia's most sophisticated Blackwell processors. Warren, the committee's ranking member, accused the administration of "reckless mismanagement" of the Bureau and said the failure to update export controls may have allowed advanced U.S. AI chips to flow to companies headquartered in China.

Read at Investing.com ↗ • Read at Senate Banking ↗

Warren opens separate Nvidia export compliance probe with June 18 response deadline

Sen. Elizabeth Warren on June 1 sent a letter to Nvidia Executive Vice President Tim Teter and Audit Committee Chair Brooke Seawell seeking answers on the company's compliance with U.S. export control laws and board-level oversight of those controls, per Bloomberg Law. The Senate Banking ranking member cited recent Justice Department enforcement actions alleging unlawful diversion of Nvidia GPUs to China through Malaysia and Thailand, $160 million in attempted H100 and H200 chip exports, and $510 million in diverted servers, per Traders Union. Warren said the criminal cases conflict with CEO Jensen Huang's public statements that there is no evidence of AI chip diversion or that Nvidia's chip market share in China has dropped to zero. Nvidia responses are requested by June 18, 2026.

Read at Senate Banking ↗ • Read at Bloomberg Law ↗ • Read at Traders Union ↗

NYT and Bloomberg document seven PLA-linked Chinese universities seeking Nvidia H200 access

A New York Times analysis of six years of People's Liberation Army procurement records, published Monday, documented systematic Chinese military efforts to acquire restricted U.S. technology including Nvidia chips. A companion Bloomberg review identified at least seven Chinese universities supporting the country's armed forces and defense industry seeking H200 access, including Beihang University and Northwestern Polytechnical University from China's "Seven Sons of National Defense" group, both already blacklisted by the Commerce Department. The earliest documented H200 leasing request dates to June 2025, and the compute lease structure exploits a regulatory gap because the hardware does not cross borders and therefore does not count as an export. American Enterprise Institute fellow Ryan Fedasiuk said the institutions are pursuing Nvidia chips "both because nobody can beat Nvidia on quality, and because there are not a sufficient number of Huawei-designed alternatives in production for them to bid on."

Read at NYT ↗ • Read at Bloomberg Law ↗

Anthropic confidentially files SEC draft S-1, edging ahead of OpenAI to public markets

Anthropic said Monday it has confidentially filed a draft S-1 with the Securities and Exchange Commission, putting the Claude maker ahead of OpenAI in the race to a U.S. public listing, CNBC reported. The company said in a statement the filing gives it the option to go public after the SEC completes its review, and that the offering will depend on market conditions; Anthropic said it has not decided on share count or pricing. Anthropic raised $65 billion last week at a $965 billion post-money valuation and is now booking $47 billion in annualized revenue from Claude, per the Associated Press. Reuters has reported OpenAI is preparing to confidentially file in the coming weeks; OpenAI CEO Sam Altman, in an interview Monday, said he is not focused on IPO timing and will go public when it makes sense.

Read at CNBC ↗ • Read at AP ↗ • Read at Anthropic ↗

OpenAI distances itself from cofounder Greg Brockman's $25M donation to Leading the Future

OpenAI on Monday published a post separating the company from political contributions made by cofounder and President Greg Brockman, saying it does not direct or have visibility into the operations of Leading the Future, the pro-AI network to which Brockman and his wife gave $25 million in 2025, Business Insider reported. OpenAI said it has not donated to any super PAC or political campaign and does not run an employee-funded PAC, calling for "thoughtful regulation" of AI and for advocacy groups to disclose their views, be honest about whom they represent, and avoid astroturfing. LTF's primary super PAC had raised more than $50 million by the end of 2025, with Andreessen Horowitz contributing $25 million, per FEC filings.

Read at Business Insider ↗

II.China Watch

Xi's Qiushi article names embodied AI and brain computer interfaces among six "future industries"

Xi Jinping's lead article in the June 1 issue of Qiushi, the Communist Party's flagship theory journal, formalizes six "future industries" as strategic priorities, per Sinocism. The list names embodied intelligence and brain computer interfaces alongside quantum, biomanufacturing, hydrogen and fusion energy, and sixth-generation mobile. The piece, drawn from a Jan. 30, 2026 Politburo Study Session, casts the priorities as "the commanding heights of science, technology, and industry." The report said the issue landed alongside the State Council's new Regulations on Outbound Investment, released June 1 with a joint Q&A from the Ministry of Justice, NDRC and MOFCOM, that consolidate prior departmental rules to address geopolitical risks and intensifying international competition.

Read at Sinocism ↗

ByteDance to begin charging for AI assistant Doubao in late June, with Douyin commerce next

ByteDance's AI assistant Doubao plans to launch paid subscription services in late June, with three tiers at 68 yuan ($9), 200 yuan ($28) and 500 yuan ($70) per month covering higher compute tasks such as presentation generation, data analysis and video production, TechNode reported, citing 36Kr. PC and mobile billing integration is expected to take about a month. The article said paid use cases will expand in the third quarter to ByteDance's Douyin e-commerce platform, with subsidies steering subscribers toward the marketplace. Doubao had about 336 million monthly active users in April, and ByteDance is accelerating monetization as compute costs rise.

Read at TechNode ↗ • Read at 36Kr ↗

Alibaba's Qwen team enters embodied AI race with vision language action model

Alibaba's Tongyi Qianwen team launched Qwen-VLA, its first vision language action model for embodied AI, extending the Qwen vision language stack with an action decoder layer designed to drive multiple robot platforms from a single set of weights, Pandaily reported. The model uses embodiment aware prompts so the same weights can control robots with different hardware and control conventions, with the team publishing benchmark gains across simulation and real world bimanual manipulation tasks. Chinese frontier labs have prioritized embodied intelligence in 2026, with the Qwen team joining DeepSeek and Tsinghua's AIR lab in publishing robot foundation models. Alibaba's entry brings China's largest open weight model family into direct competition for robotics developers.

Read at Pandaily ↗

WeRide, Uber and AVOMO announce Madrid robotaxi pilot, Spain's first

Chinese self-driving company WeRide, Uber and fleet operator AVOMO announced plans to launch Spain's first commercial robotaxi pilot in the Region of Madrid, with the service expected to go live later in 2026, Jiemian reported. WeRide CEO Tony Han called Spain the company's fifth European market entry. The Madrid pilot is the fourth city under WeRide's 15 city Uber agreement, building on Middle East routes where fully driverless commercial service is already running. Rides will be available through the Uber app, with safety operators on board at the start.

Read at Jiemian ↗ • Read at WeRide ↗

III.Federal Policy Tracker

Sanders to introduce bill taxing AI firms' stock to fund a public sovereign wealth fund

Sen. Bernie Sanders (I-Vt.) wrote in a New York Times op-ed Monday that he will soon introduce the American A.I. Sovereign Wealth Fund Act, which would give the public an ownership stake in the largest U.S. AI companies through a one-time 50 percent tax paid not in profits but in stock from firms including OpenAI, Anthropic and xAI. Sanders argued that because AI is built on humanity's collective knowledge, the wealth it generates should benefit the public, and said the government would gain voting shares and equal board representation to block decisions that harm citizens. Proceeds would fund direct payments to Americans alongside health care, education and housing. He noted that OpenAI and Anthropic have each floated public or sovereign wealth fund concepts and cited Norway's and Alaska's oil-funded models as precedent, with full legislative details due in the coming weeks.

Read at NYT ↗

Warren and Grassley introduce bipartisan defense contractor foreign ownership disclosure bill

Sens. Elizabeth Warren (D-Mass.) and Chuck Grassley (R-Iowa) on Monday introduced bipartisan legislation that would require defense contractors and subcontractors to disclose beneficial owners and any foreign ownership, per Semafor. The senators said they will seek to add the measure to the FY27 National Defense Authorization Act. The bill lowers the disclosure threshold for foreign ownership, control or influence to Pentagon contracts above $500,000, down from the current $5 million floor. Grassley framed the proposal as bringing transparency to foreign influence in Defense Department contracts, and Warren described it as commonsense bipartisan reform to clarify with whom the Pentagon does business with. The Pentagon declined to comment on the proposed legislation.

Read at Senate Banking ↗ • Read at Semafor ↗

VA opens RFI for AI interface and LLM API services covering its workforce

The Department of Veterans Affairs updated a request for information June 1 seeking input on AI user interface and application programming interface services that would provide assistive, collaborative and agentic AI to the VA workforce, FedScoop reported. The RFI was originally posted Friday, May 29, and revised Monday. The notice covers VA wide deployment of large language model access for employees.

Read at FedScoop ↗

IV.Capability & Research Watch

Researchers exploit Meta AI support bot to hijack high profile Instagram accounts including Obama White House handle

Meta confirmed Monday that hackers used the company's AI powered customer support chatbot to take over high profile Instagram accounts including Barack Obama's White House account, Sephora and the U.S. Space Force chief master sergeant's account, The Guardian reported, citing 404 Media. Ordinary users on Reddit and X reported similar account hijackings over the weekend. The attack told Meta's AI assistant to link the target account to a new email address; the bot then sent a verification code to the attacker's email, which the attacker pasted back into the chat interface to trigger a password reset on the targeted account. In at least one demonstration shared on social media, the attacker used a VPN to spoof the account holder's location and avoid Meta's safeguards, and stolen Instagram handles were listed for sale on Telegram. Meta said in a statement the issue has been resolved and that it is securing impacted accounts; the company did not say how many accounts were affected.

Read at The Guardian ↗

LessWrong analysis argues die-level compute verification is not feasible by 2028 and urges board and firmware paths

A LessWrong post published Tuesday argues that compute verification for AI export control purposes cannot work at the chip die level on short timelines, because semiconductor design lead times put die-level changes more than two years out. The post estimates board-level changes take months and software or firmware changes one to two months. That leaves Nvidia's Rubin generation chips, expected later this year, accessible only to software-level verification, and Feynman generation chips expected in 2028 accessible to board-level changes. The author proposes a board-level microcontroller that hashes the weights sent to the die and keeps a dictionary of hashes, with a small number of distinct hashes implying no weight updates and many hashes implying updates, building on prior work by Shavit (2023).

Read at LessWrong ↗

V.Industry & Market Watch

Alphabet raises $80B for AI buildout with $10B Berkshire Hathaway anchor stake

Alphabet on Monday announced plans to raise $80 billion in equity to fund its AI infrastructure, including a $10 billion private placement with Berkshire Hathaway, the Wall Street Journal reported. The financing structure runs $30 billion in underwritten public offerings split between mandatory convertible preferred stock and common stock, $40 billion through an at-the-market stock offering program expected to begin in the third quarter, and the $10 billion Berkshire private placement, per Axios. Berkshire will buy $5 billion of Class A common stock at $351.81 per share and $5 billion of Class C capital stock at $348.20 per share, expanding a position the conglomerate has been building since the fall of 2025 to nearly $17 billion as of March, per the Associated Press. Alphabet said the proceeds support capital expenditures to scale AI infrastructure and global compute amid customer demand exceeding the company's available supply.

Read at WSJ ↗ • Read at Axios ↗ • Read at AP ↗

Nvidia unveils RTX Spark superchip for AI agent PCs with Microsoft, Dell, HP and Lenovo

Nvidia CEO Jensen Huang on Monday unveiled the RTX Spark superchip combining CPU and GPU capabilities for laptops and desktops designed to run AI agents locally, with models expected to ship in the fall from Microsoft, Dell, HP, Lenovo, Asus and MSI, the Wall Street Journal reported. The 1 petaflop chip is paired with secure sandboxes jointly developed with Microsoft to run agents securely on device, with more than 100 Windows software makers including Adobe, Blender and ComfyUI signed on to support the new chip, per TechCrunch. Microsoft branded its own RTX Spark laptop the Surface Laptop Ultra, calling it the most powerful Surface Laptop the company has built. Huang told investors last month that Nvidia has identified a $200 billion CPU market opportunity for AI in addition to its existing GPU business, and that the company has already sold $20 billion of its earlier Vera server CPU.

Read at WSJ ↗ • Read at TechCrunch ↗ • Read at AP ↗

Newsom-aligned foundations hold small Anthropic stake as company files for IPO

Three mission-focused foundations, the Ford Foundation ($5 million), Omidyar Network ($1.5 million), and the Nathan Cummings Foundation ($1 million), together committed $7.5 million to buy just under 250,000 Anthropic shares in the 2024 sale of FTX's bankrupt estate, less than 1% of that $884 million auction, ImpactAlpha reported. Omidyar and Ford have funded causes promoted by California Gov. Gavin Newsom, and the New York Post estimated the combined stake could be worth roughly $250 million at Anthropic's $965 billion valuation, the level at which Anthropic filed confidentially for an IPO on June 1. Anthropic products are deployed in a California Department of Tax and Fee Administration customer-service tool and in Newsom's "Engaged California" AI consultation, and Newsom signed Anthropic-backed Senate Bill 53, the state's frontier-AI transparency law, in 2025. Omidyar's lead on the deal said the foundations are "under no illusions" about their limited influence. Newsom spokesperson Tara Gallegos called the framing a "conspiracy theory."

Read at NY Post ↗ • Read at ImpactAlpha ↗