AI Policy · Daily

New York paused new hyperscale data centers above 50 megawatts under a Hochul executive order, giving state utility regulators up to a year to write standards on water, air and rate impacts. Xi Jinping will deliver the keynote at the World Artificial Intelligence Conference in Shanghai on Friday, the first time China's president has headlined the summit. Roughly 200 economists including 15 Nobel laureates signed a "We Must Act Now" statement organized by Stanford's Erik Brynjolfsson urging faster action on AI's labor and economic effects. Three major publishers and author Scott Turow sued Google in a Southern District of New York class action alleging its Gemini models were trained on books lifted from Google Books, Play Books and Scholar.

I.Top Stories

Hochul enacts New York moratorium on new hyperscale data centers over 50 megawatts

Gov. Kathy Hochul (D) signed an executive order blocking new environmental permits for data centers over 50 megawatts for up to a year, per the Verge. The order gives state regulators time to write standards on water use, air quality and rate impact, directs the Department of Environmental Conservation to withhold discretionary permits until a generic environmental review is complete, and the governor's office said the 50 MW threshold is meant to spare smaller facilities used by institutions like hospitals. That threshold is higher than the 20 MW cap in a bill already passed by the state legislature, which still awaits Hochul's signature. Hochul said the pause is her responsibility given data centers' effect on utility bills, natural resources and uncertainty for New Yorkers.

Read at The Verge ↗ • Read at AP ↗

Nearly 200 economists including 16 Nobel laureates urge policymakers to move faster on AI risk

A statement titled "We Must Act Now," released Monday and organized by Stanford economist Erik Brynjolfsson with Ajay Agrawal, Anton Korinek and Tom Cunningham, warned that AI's effects could be "larger than the Industrial Revolution, but unfolding over a vastly shorter time frame," the New York Times reported. Signatories include 16 Nobel laureates, the chief economists of OpenAI and Anthropic, Anthropic co-founder Jack Clark, former Google CEO Eric Schmidt and venture capitalist Vinod Khosla. The list includes MIT economists Daron Acemoglu and Simon Johnson, the 2024 Nobel laureates who had previously been among the profession's more prominent AI skeptics. The statement calls on economists, policymakers and industry leaders to act now on AI's economic effects. Brynjolfsson said "I still see a big gap there, a big mismatch, and I'm kind of worried that we're not going to be ready for the tsunami that's coming."

Read at New York Times ↗ • Read at Stanford Digital Economy Lab ↗

White House to convene utilities, data center developers and governors to expand voluntary Ratepayer Protection Pledge

The Trump administration plans an event "in the coming weeks" bringing utilities, companies that build and operate data centers for Big Tech, and governors of states leading power infrastructure expansion into an enlarged voluntary pledge, per Reuters. Earlier signatories at the initial White House ceremony this year include Amazon, Google, Meta, Microsoft, OpenAI, Oracle and xAI; the original pledge commits companies to finance new power generation, grid upgrades and unused reserved capacity tied to their AI projects rather than pass those costs to existing utility customers. A White House official said additional stakeholders want to sign the pledge because the original has been so impactful.

Read at Reuters ↗ • Read at The White House ↗

Rep. Casar calls on Democrats to refuse AI industry PAC money, unveils counter-package

Rep. Greg Casar (D-Texas), chair of the Congressional Progressive Caucus, told Politico that Democrats are "in the wilderness" on AI and must stop taking money from AI industry super PACs, singling out Leading the Future's $8 million spend to defeat an industry critic in a New York City House primary. His platform calls for banning AI driven personalized pricing from consumer data, requiring data centers to generate their own electricity, and imposing chatbot guardrails; Casar is separately introducing legislation to tax AI "tokens" (the underlying unit of AI computing power) to slow the industry's growth. That framework goes beyond Minority Leader Hakeem Jeffries and House Democratic leaders, who have kept the party's AI message focused on shielding ratepayers from data center costs and blocking the Trump administration from writing federal AI rules. Casar said the three-member commission Jeffries convened to develop a consensus Democratic AI framework does not expect to introduce any legislation until year-end, after the House majority has likely been decided.

Read at Politico ↗

Hachette, Cengage, Elsevier and Scott Turow file class action against Google over Gemini training

The three publishers and author Scott Turow filed a class action complaint July 13 in the Southern District of New York, alleging Google copied books it obtained through Google Books, Google Play Books and Google Scholar to train early Gemini models, per Publishers Weekly. The complaint cites an internal Google document estimating potential fines in the tens to hundreds of billions of dollars, and asserts Google went ahead knowing publishers would consider the practice illegal. Plaintiffs seek statutory damages, a permanent injunction against further infringement and destruction of unauthorized copies. Hachette and Cengage had earlier moved to join the 2023 Google Generative AI Copyright Litigation brought by illustrators and writers, but withdrew that motion to bring this class-wide complaint after concluding a three-year statute of limitations could bar many of their claims inside the older case.

Read at Publishers Weekly ↗

II.China Watch

Xi Jinping will keynote China's flagship AI summit for the first time in a bid to shape global AI rules

Sinocism reported Monday that President Xi Jinping will deliver the keynote at the World Artificial Intelligence Conference (WAIC) in Shanghai on Friday, an upgrade from prior years when Premier Li Qiang was the most senior speaker. Citing a Xinhua (state media) commentary, the newsletter said Beijing plans to use the platform to promote its Global AI Governance Initiative, its proposal to establish a World AI Cooperation Organization, and its push for an open source, open ecosystem model of technology sharing with developing countries. Organizers said the July 17 to 20 event will span three Shanghai zones and draw more than 1,100 exhibitors alongside 1,400 international guests. Xi's personal appearance turns WAIC from an industry showcase into a state level venue for governance diplomacy and lands as Beijing continues to press UN General Assembly resolutions on AI cooperation that would give it a lead role in shaping multilateral rules.

Read at Sinocism ↗

China's internet industry adopts a self-regulatory pact on how AI agents handle personal data

The China Internet Association released a self-regulatory pact on personal information protection for AI agents at a Beijing forum, with Baidu, Tencent, Alibaba, Volcengine and 27 other companies as first signatories, TechNode reported. The pact aims to standardize how agent services collect, process and use personal data as those services spread across major platforms. The association separately released a pact for mini-program ecosystems signed by Tencent, Ant Group, Baidu and other operators. The move lands two days before the Cyberspace Administration's Interim Measures on anthropomorphic AI interaction services take effect on July 15.

Read at TechNode ↗

StepFun launches an agent-native operating system and the first phone built around it

Chinese AI startup StepFun on Monday unveiled Step AOS, an agent-native operating system, along with the STEPX Neo phone that runs on it, Caixin reported. Step AOS routes tasks across Alipay, Meituan, JD.com, Didi, Ctrip, Amap and Baidu through their protocol interfaces rather than the graphical user interface, letting an on-device agent called Amoo string together end-to-end actions such as ordering, payment and navigation. Chairman Yin Qi said at the launch that China's large language model business path may not follow the AI coding route Western labs are pursuing, and that software-hardware integration is the direction StepFun is willing to bet on. The launch pushes agent-driven commerce into consumer hardware and previews the interconnection protocols Chinese platforms will need to expose to comply with Beijing's forthcoming agent governance framework.

Read at Caixin ↗

SenseTime open sources a unified vision model designed to replace multiple task specific systems

SenseTime released and fully open sourced SenseNova-Vision, a unified vision foundation model built to handle object detection, optical character recognition, image segmentation, depth estimation and 3D reconstruction in one system, TechNode reported. The company also published SenseNova-Vision Corpus-50M, a 50 million sample visual instruction dataset, and said the new model will be folded into its SenseNova U-series under an Apache 2.0 license. SenseTime has been on the U.S. Commerce Department's Entity List since 2019, restricting its access to advanced U.S. chips and design tools. The release extends China's open weight strategy from language into vision and gives Chinese robotics and autonomous driving developers a domestically controlled vision stack.

Read at TechNode ↗

III.Policy Tracker

HUD withholds DOGE AI documents citing an AI privilege Wired says does not exist

The Department of Housing and Urban Development declined to release more than 100 documents in response to a Democracy Forward FOIA request about DOGE's use of AI at the agency to identify federal regulations for potential rescission, per Wired. Withholding reasons cited included "deliberative AI input," "draft of AI prompt" and a "presidential communications privilege" that is generally held to apply only to the president and immediate advisers. Named documents belonged to former DOGE HUD staffers Christopher Sweet, a June University of Chicago graduate, and Scott Langmack, now executive director of deregulation AI at the Office of Management and Budget. Tori Noble, a staff attorney at the Electronic Frontier Foundation, told Wired that access to the prompts is needed to check whether AI tools "hallucinate, show bias, or just plain get things wrong." No federal law currently requires agencies to disclose when AI has been used in the drafting or rescission of rules.

Read at Wired ↗

IV.Industry & Market Watch

Georgia Power acquires more than 300 parcels for AI data center transmission line as homeowners describe forced sales

Georgia Power is acquiring more than 300 land parcels for a new transmission line where the utility estimates 70% to 80% of the electricity will serve data centers and the remaining 20% to 30% will serve growing residential and commercial demand, CBS News reported. Ansley Brown, a homeowner whose mother agreed to sell rather than face eminent domain proceedings, told CBS "to us it's theft. It's literally a billion dollar company stealing land from smaller people, people who can't fight back." Georgia Power spokesperson Holly Lovett said eminent domain is always a last resort for the company. Brown has been aggregating stories from other displaced Georgia families on TikTok.

Read at CBS News ↗

Google DeepMind's Hassabis calls for a U.S.-led global AI watchdog

Demis Hassabis, Google DeepMind co-founder and CEO, is urging the U.S. to create a new AI standards body empowered to screen the world's most advanced models and coordinate an industry-wide slowdown if dangers mount, Axios reported Tuesday. The body would be modeled on FINRA — industry-funded, staffed by technical experts and answerable to the U.S. government — with frontier labs initially sharing models voluntarily up to 30 days before release for testing of dangerous cyber, biological and "deception" capabilities, and mandatory pre-deployment clearance once the regime proves effective. The rules would cover all frontier-class models regardless of country of origin or whether they are open or closed, and would be overseen by a majority-independent board. Hassabis, who has briefed the Trump administration, rival lab leaders and European officials, wants the body running before year-end.

Read at Axios ↗ • Read at Demis Hassabis ↗

Data center developers pitch majority equity stakes worth tens of billions to private equity as AI demand crests

Bankers are working to sell majority stakes in U.S. data center builders and operators including Netrality Data Centers, DataBank, Edged and EdgeCore Digital Infrastructure to private equity firms this summer, per WSJ. The properties span Phoenix to Atlanta and are being pitched on unrelenting demand for computing power to run AI workloads. A backlog of server capacity demand has driven data center operators to novel strategies including renting chips from direct competitors and launching data centers into orbit.

Read at WSJ ↗ • Read at Mint ↗

Meta ramps up AI smartglasses marketing while testing 'super-sensing' prototype

Meta is aggressively marketing its Ray-Ban AI smartglasses, which capture what the wearer sees and hears, even as it tests a prototype that continuously collects audio and takes photos every few seconds, per WSJ. Meta updated the current glasses so the camera shuts off if the LED recording light is tampered with or destroyed, and said it may pursue legal action against sellers offering to disable the LED, per Fortune. The company is facing a lawsuit alleging intimate moments captured by the glasses were later reviewed by Kenya based contractors training Meta's AI models. Executives have discussed not activating the LED while super-sensing features are in use, per Financial Times reporting.

Read at WSJ ↗ • Read at Fortune ↗ • Read at TechCrunch ↗

Google signs 1.6 GW Arkansas solar PPA to offset emissions from AI data centers

Google agreed to buy all of the initial electricity output from the Steel River Energy Center in Arkansas under a power purchase agreement (PPA), coming online in 2029 with 1.6 gigawatts of solar and 2 gigawatt hours of battery storage, per FT. The deal shows hyperscalers continuing to buy U.S. renewables even as the Trump administration works to end Inflation Reduction Act clean energy tax credits and stall project plans. The facility is designed to expand to 2.5 gigawatts of solar and 2.9 gigawatt hours of storage at full completion. Google will pay a fixed price for the electricity without physically receiving the energy.

Read at FT ↗ • Read at Investing.com ↗

Montefiore lays off 12 utilization review nurses in the Bronx after AI software takeover, union says

Twelve nurses at Montefiore Medical Center in the Bronx, whose jobs involved reading patient charts and negotiating with insurers over coverage, were laid off Sunday after being replaced by AI powered software, the New York State Nurses Association told The Guardian. Marilyn Shuler, a utilization review nurse of 39 years, was among those laid off. Union executive committee member Shaiju Kalathil said the layoffs violate the contract nurses secured in the January 2026 New York City hospital strike, which included AI related safeguards. National Nurses United president Jamie Brown said the parent union has developed an AI bill of rights for patients and nurses and has been pushing for guardrails through contracts and legislation.

Read at The Guardian ↗

V.Global & Geopolitics

Intel commits €5 billion ($5.7 billion) to Irish fab expansion for AI driven Xeon production

Intel began a €5 billion ($5.7 billion) capital investment to upgrade and expand its Leixlip campus outside Dublin, adding capacity to produce Intel Xeon 6 processors and next generation Xeon chips built on the Intel 3 manufacturing process, per Financial Times. Intel Foundry Executive Vice President Naga Chandrasekaran said the build-out represents about 30% of Intel's $17 billion planned 2026 capital expenditure and will add several hundred jobs to the roughly 4,900 people Intel employs in Ireland. Taoiseach Micheál Martin, Ireland's prime minister, called the expansion a vote of confidence in Ireland's advanced manufacturing base, per the Irish Examiner.

Read at FT ↗ • Read at Reuters ↗ • Read at Irish Examiner ↗

Finnish AI lab NestAI launches operating engine for Finnish and Estonian militaries

NestAI Oy on Monday announced an AI engine that Finnish and Estonian militaries will use to build sovereign military models for operational planning and unmanned autonomous systems, per Bloomberg Technology. The two governments and the company signed a letter of intent June 30 through the Finnish Defense Forces AI Center of Excellence and Estonia's Force Transformation Command, per AeroTime; the framework carries no financial commitment and pools research, development, training and technical cooperation. Both sides described the effort as built on open, modular architectures to interoperate with allies while retaining sovereign control of data, technology and operational application. NestAI raised €100 million (about $110 million) from Nokia and Finnish state fund Tesi in late 2025; founder Peter Sarlin previously sold Finnish AI company Silo AI to AMD.

Read at Bloomberg ↗ • Read at AeroTime ↗

Ireland's Dáil debates Sinn Féin AI copyright motion as artists lobby at Leinster House

A Private Members' motion from opposition party Sinn Féin lawmaker Aengus Ó Snodaigh calling on the government to protect the copyright of Irish artists in the era of artificial intelligence is scheduled for debate in the Dáil, Ireland's lower house of parliament, on the evening of July 14, per RTÉ. Backers include the Irish Music Rights Organisation, the Screen Composers Guild of Ireland and the Ivors Academy, alongside artists Danny O'Donoghue of The Script and Tom Dunne of Something Happens. O'Donoghue urged the government to back the motion to protect Irish songwriters and composers, saying creators deserve "their consent, fair pay and transparency." The motion also urges Ireland to use its EU Council presidency to lead European action on creators' rights.

Read at RTÉ ↗