Pentagon bars Alibaba, Baidu, BYD from defense contracts | Schiff intros Pentagon AI bill | OpenAI files for IPO

Bottom Line Up Front

The Pentagon added Alibaba, Baidu and BYD to its roster of Chinese military companies barred from U.S. defense contracts, expanding the list to 188 entities. Sen. Adam Schiff filed the Human Authority in Lethal Operations Act to require human command over Pentagon autonomous weapons, landing on the eve of Senate Armed Services FY27 NDAA markups. The Office of Personnel Management enrolled Cisco, Scale AI, Wiz and Cognition AI in its Tech Force partner program as a 60-day deadline runs on President Trump's June 2 AI cybersecurity executive order.

I.Top Stories

Pentagon bars Alibaba, Baidu and BYD from defense contracts

The Pentagon on Monday updated its Section 1260H list of Chinese military companies, adding Alibaba, Baidu and BYD and growing the roster to 188 entities from roughly 130 last year, per the Associated Press. Designation blocks the named companies from receiving U.S. defense contracts and may cascade to private sector counterparty risk. Alibaba and Baidu both rejected the designation as baseless, and the Chinese Embassy in Washington accused Washington of overstretching national security and discriminating against Chinese firms. The House Select Committee on the CCP called the list a warning to American businesses, all levels of government and the American people.

Read at NPR ↗

Sen. Schiff introduces Human Authority in Lethal Operations Act ahead of Senate NDAA markup

Sen. Adam Schiff (D-Calif.) on Monday introduced the Human Authority in Lethal Operations (HALO) Act, which would require human involvement in Pentagon use of autonomous or semi-autonomous AI enabled weapons, The Hill reported. The bill establishes a chain of command under which a designated commander has the final say over decisions involving such systems. HALO is the newest addition to a recent uptick of Pentagon-related AI measures, including standalone bills and House NDAA amendments. Senate Armed Services begins multiday FY27 NDAA subcommittee and full-committee markups Tuesday.

Read at The Hill ↗ • Read at Senator Schiff ↗

OPM adds Cisco, Scale AI, Wiz and Cognition AI to Tech Force as 60-day clock on Trump AI cybersecurity EO runs

The Office of Personnel Management on Monday added Cisco, Scale AI, Wiz, Arista Networks, Armada, Cognition AI, Cognizant, Payward and Moveworks to its Tech Force industry partner roster, FedScoop reported. They join OpenAI, Google Public Sector, xAI and Palantir among the program's two dozen existing partners. Partners provide training resources and nominate their own employees for temporary federal service, with the program aiming to bring on roughly 1,000 early career hires on two year terms; it has made about 200 hires and onboarded around 10 to date. President Trump's June 2 AI cybersecurity executive order directs Tech Force to expand AI and cyber placement within 60 days, and a separate National Security Presidential Memorandum tells OPM to use special hiring and pay authorities for AI talent.

Read at FedScoop ↗

Sacks tells Fox Business an 'FDA for AI' would cede the race to China, citing a six- to nine-month U.S. lead

Former White House "AI czar" David Sacks said on Fox Business "Kudlow" Monday that an "FDA for AI" regulatory approach could hand the AI race to China, putting the U.S. lead at only six to nine months. Sacks likened Washington's regulatory impulse to climate politics, telling host Larry Kudlow that "AI has become the new climate change." He acknowledged that some frontier models, including Anthropic's Mythos, raise cybersecurity concerns but cautioned against a moral panic over the technology. The remarks come days after President Trump signed the June 2 voluntary AI cybersecurity executive order.

Read at Fox Business ↗

Harvard professors counter Bernie Sanders' AI sovereign wealth fund with energy tax pitch

In a Guardian op-ed published Monday, Nathan E. Sanders and Bruce Schneier proposed taxing AI companies' energy use as an alternative to Sen. Bernie Sanders' American AI Sovereign Wealth Fund Act. The authors say energy taxation gives policymakers a direct way to make AI companies pay for the social costs of their compute footprint while preserving the senator's goal of curbing tech oligarch concentration. They do not oppose a 50% public equity stake in firms such as Anthropic, OpenAI and xAI but say behavior-shifting taxes on power demand are the stronger lever. The piece extends arguments from the authors' book, "Rewiring Democracy."

Read at The Guardian ↗

II.China Watch

China's state security ministry warns on third-party 'AI transfer station' proxy services

China's Ministry of State Security on Monday published a piece on its WeChat public account titled "'AI Transfer Stations': Guard Against the Risks," Sinocism reported. The MSS said the fast-growing market for third party services consolidating multiple AI model APIs has produced widespread privacy leaks, resale of user prompts to other vendors for training, model substitution, embedded backdoors and unauthorized cross-border data transfers. Per Sinocism, the MSS did not flag that the most common reason Chinese users pay for these transfer stations is to reach frontier models from OpenAI, Anthropic and Google blocked or unavailable through licensed PRC channels, even as the MSS's own list of platform appeals includes "bypassing usage restrictions."

Read at Sinocism ↗

MIIT and SASAC order central SOEs and provinces to anchor humanoid robot deployment

China's Ministry of Industry and Information Technology and the State-owned Assets Supervision and Administration Commission have jointly issued a notice launching the 2026 Humanoid Robot and Embodied AI Real-Scene Training Special Action, Pandaily reported. The plan sets a target of identifying more than 100 high value application scenarios and establishing thousand-unit-scale deployment by year-end. Provincial authorities must select at least 20 key scenarios across industrial, service or specialized sectors, and central state-owned enterprises must identify at least 10 in their industries. The framework organizes end users, robot makers, algorithm developers and supply chain firms into "Innovation Application Consortia" for each scenario.

Read at Pandaily ↗

Alibaba in talks with Chinese nuclear SOE on small modular reactor for Hangzhou data center

Alibaba has approached a Chinese state-owned nuclear power enterprise about building small modular reactors to supply its Renhe Data Center in Hangzhou, Pandaily reported, citing an industry source quoted by China Business Journal. The facility provides computing infrastructure for AI customers including BrainCo Technology and Lingban Technology, and faces growing electricity demand from AI workloads. The talks have stalled over China's unified state grid system, which requires SMR-generated electricity to be fed through the public grid at roughly 0.58 yuan ($0.08) per kilowatt-hour, eroding the cost advantage of a dedicated nuclear feed. The nuclear enterprise source said a self-supply model would be needed for the project to be economical, but current policy does not allow it.

Read at Pandaily ↗

Tencent opens WeChat AI Agent developer guidelines with Meituan, Ctrip and Tongcheng on board

Tencent's WeChat Open Platform released developer integration guidelines for the long-rumored WeChat AI, 36Kr reported. Mini-programs that integrate gain the ability to be recommended and invoked by WeChat's coming agent; those that do not integrate cannot be invoked. Meituan announced same-day integration as the first internal test partner, with Ctrip and Tongcheng following soon after. Tencent customer service had earlier said WeChat is also working with Huawei, Xiaomi, Honor, OPPO and vivo on agent-to-agent capability that lets the AI assistants in those handset systems initiate WeChat audio, video or messaging.

Read at 36Kr ↗

III.Federal Policy Tracker

Senate Armed Services opens FY27 NDAA markup with AI provisions sitting inside Emerging Threats subcommittee jurisdiction

The Senate Armed Services Committee opens multiday FY27 NDAA markups Tuesday, with the full committee closed session at 2:30 p.m. in Russell 222 and subcommittees running through the day, per Congress.gov. Closed subcommittee markups include Cybersecurity at 11:00 a.m. and Emerging Threats and Capabilities at 9:15 a.m., which holds jurisdiction over AI, autonomy, biotech and quantum. The Coons-Reed Responsible AI in Defense Act, Gillibrand's Secure and Accountable Military AI Act and Schiff's HALO Act are all candidates for inclusion as amendments to the base bill.

Read at Congress.gov ↗ • Read at U.S. Senate Committee on Armed Services ↗

Federal judge in Boston strikes down $100,000 H-1B fee, opening circuit split

U.S. District Judge Leo Sorokin in Boston on Monday struck down the Trump administration's $100,000 fee on new H-1B visas, siding with 20 states, per the Associated Press. Sorokin wrote that the policy "imposes a tax on H-1B petitions without the requisite delegation by Congress" and violated the Administrative Procedure Act. The summary judgment is the opposite of an earlier federal court ruling that upheld the same fee, and a third lawsuit by religious groups and labor organizations is pending in San Francisco, setting up the possibility of divided rulings across three appellate circuits. H-1B users are concentrated in tech, with nearly three-quarters of approvals going to workers from India.

Read at NPR ↗

House Science holds DOE FY27 budget hearing Wednesday covering Office of Science AI for Science and ASCR

The House Science, Space, and Technology Committee holds a full-committee hearing Wednesday at 10:00 a.m. in Rayburn 2318 on the Department of Energy's FY27 budget request, per the committee. DOE houses the Office of Science AI for Science initiative, the Advanced Scientific Computing Research program and the national lab AI infrastructure stack. DOE's FY27 budget justification documents are public and detail allocations across the department's national laboratories.

Read at Congress.gov ↗ • Read at House Committee on Science, Space & Technology Republicans ↗

Warren says Senate Banking 'AI and the American Dream' hearing will proceed without Nvidia CEO

Sen. Elizabeth Warren (D-Mass.), Banking ranking member, released a statement Monday after Nvidia President and CEO Jensen Huang declined her invitation to testify at the committee's June 11 hearing on AI. Warren said Huang owed Congress public answers given that he had "time to attend a $1 million-a-head dinner at Mar-a-Lago and fly across the world to meet with President Xi Jinping of China." Huang offered to host committee members at Nvidia's Santa Clara headquarters for private discussions, per CNBC. Warren has asked Huang to testify about Nvidia's views on U.S. export controls and its business in China, with Banking minority staff in contact with the company on alternative open session dates.

Read at Senate Banking ↗ • Read at CNBC ↗

IV.Capability & Research Watch

Stanford study finds AI hiring tool monoculture screens out 26% of Black candidates across 156 employers

A Stanford HAI study described as the first large scale analysis of hiring algorithms in the wild found AI screening tools rejected nearly 26% of Black candidates across more than 4 million applications from 3.4 million people, 156 employers and 1,700 job postings, all using the same AI hiring vendor, The Root reported. Recommending Black and Asian candidates at the same rate as the most favored group, typically white applicants, would have advanced 40,000 more applications to the next stage, the researchers found. The algorithm was less likely to recommend Black candidates for finance roles but more likely to recommend them for warehouse roles, with the offsetting patterns concealing the discrimination. More than 90% of companies globally use some form of AI for initial candidate screening, per a World Economic Forum statistic cited in the study.

Read at The Root ↗ • Read at Stanford Digital Economy Lab ↗

Apple builds Google Cloud and Nvidia into Private Cloud Compute for Siri AI ahead of iOS 27

Apple at WWDC 2026 Monday disclosed it is expanding Private Cloud Compute beyond its own data centers to run Apple Intelligence workloads on Google Cloud and to use Nvidia chips, MacRumors reported. Apple Foundation Models will incorporate technology behind Google's Gemini, with some Apple Intelligence processing offloaded to Google servers. The infrastructure expansion accompanies Apple's Siri AI reveal under the Apple Intelligence umbrella, with a wide release scheduled for fall 2026 alongside iOS 27.

Read at MacRumors ↗ • Read at TechCrunch ↗

V.Industry & Market Watch

Apollo and Blackstone close $35B chip financing package for Anthropic, structured across three tranches

Apollo Global Management and Blackstone finalized a $35 billion financing package for Anthropic to fund AI infrastructure, Investing.com reported, citing Bloomberg. The financing will fund Google custom tensor processing units that Anthropic will lease to expand its AI computing capacity. The debt package is structured across three tranches and includes backing from Broadcom. The Financial Times also reported the deal, which lands the same day as OpenAI's confidential S-1 IPO filing.

Read at Investing.com ↗ • Read at Financial Times ↗

OpenAI files confidential S-1 with SEC at $852 billion post-money valuation

OpenAI on Monday submitted a confidential draft registration statement to the SEC for a proposed IPO, the company said in a blog post, with its last post-money valuation at $852 billion. OpenAI said it published the filing voluntarily because it expected the news to leak and that the timing remains undecided and may be a long way off. The filing follows Anthropic's confidential filing the prior week, with SpaceX also expected to debut at a $1.75 trillion valuation. TechCrunch noted the SEC under the Trump administration has taken a "markedly more hands-off posture" toward tech and AI companies than under prior administrations.

Read at TechCrunch ↗ • Read at OpenAI ↗

Anduril CEO Schimpf says U.S. burned roughly 850 Tomahawks in four weeks against Iran, replenished at 90 per year

Anduril CEO Brian Schimpf told Fortune Brainstorm Tech in Aspen Monday that the U.S. and Israel ran "ten times as many strikes" in the first month of the Iran war as the entire Gulf War. The U.S. fired roughly 850 Tomahawk missiles in four weeks against a Pentagon replenishment rate of about 90 per year. Schimpf said Anduril is pursuing upstream supply security, including for germanium, citing China's acquisitions of rare earth magnets and copper film suppliers as a strategic stranglehold the U.S. has been slow to counter. He called the defense tech valuation environment a bubble, noting some companies raising at 50x to 100x forward revenue. He said the top two players in any hot category capture roughly 90% of returns.

Read at Fortune ↗

Meta removes NameTag face recognition code from Meta AI smart glasses app after Wired report

Meta removed unactivated face recognition components from the Meta AI companion app for its smart glasses one day after Wired identified the code embedded in an app installed on more than 50 million phones, Wired said in a follow-up analysis. The code, internally called NameTag, was designed to convert faces captured by Meta's glasses into on-device biometric signatures and compare them against a local database, with faces the system failed to recognize cropped, indexed and stored locally for future processing. Andy Stone, Meta's VP of communications, said the feature was purely exploratory and no final decision had been made. The New York Times reported in February that internal Meta documents weighed a launch as soon as this year, with one memo reportedly stating the release timing should fall during a politically distracted moment for privacy advocates.

Read at Wired ↗

VI.Global & Geopolitics

Apple delays Siri AI for EU iPhone users, says regulators 'refusing to engage' on DMA compliance

Apple said Monday it cannot currently launch Siri AI on iPhones, Apple Watches or iPads in the European Union because EU regulators are "refusing to engage" on its Digital Markets Act compliance proposals, per Bloomberg. Apple said it had proposed an EU specific solution that the European Commission did not accept over several months of discussions, per Business Standard. Siri AI and the new Apple Intelligence will therefore not arrive in the EU with iOS 27 when it ships in September, per AppleInsider.

Read at Business Standard ↗ • Read at AppleInsider ↗ • Read at Bloomberg ↗

UK government commits $1.47B to national AI supercomputer with procurement priority for British chip startups

The UK government on Monday detailed a $1.47 billion plan to reduce dependence on foreign made AI hardware, including more than $1 billion for a national AI supercomputer, Wired reported. The facility will be stocked with $530 million in hardware, including $200 million in specialist inference chips, with procurement priority for British firms; the government named Olix and Fractile as potential beneficiaries. British researchers and startups are expected to gain access starting in 2030. The supercomputer plan builds on April's $675 million SovAI venture fund and last November's AI growth zones initiative for permitting data centers.

Read at Wired ↗ • Read at GOV.UK ↗