The Data Center Backlash
12 GW announced. 5 GW under construction. What happens next? The Gap Between the Press Release and the Power Grid
Abbott calls for sweeping data center rules in Texas, including an end to tax breaks. A notable political shift in the state most associated with the AI buildout. The governor wants restrictions on data centers and a possible repeal of their sales tax exemption, reframing the debate around who pays and who benefits.
Over 500 groups ask Congress for a national moratorium on new data centers. The coalition wants construction paused until rules exist to protect communities and the environment. The campaign launched after the EPA said it would not set national requirements for the facilities.
East Texas residents revolt against a proposed data center. Dozens packed a county commissioners meeting to demand answers about the AmpZ Champion project, citing worries about water, noise, light, and property values. A useful case study in how AI policy is now fought at the county level.
Heartland protests push lawmakers to regulate data centers. Ground-level reporting on Midwest resistance, including a county that passed a one-year moratorium to write zoning standards. Five states, Michigan among them, are weighing full repeal of their data center incentives.
A database of 700+ policies shows US data center governance turning hostile. The analysis argues the policy mood has flipped from tax incentives toward scrutiny, moratoriums, and local restrictions, mapping how communities are responding to AI’s energy and land-use footprint.
Environmental lawsuits emerge as a real constraint on the AI buildout. Litigation over water use, air pollution, and environmental justice is slowing data center projects. Some facilities consume hundreds of millions of gallons of water a year, far more than the towns around them.
Utah residents sue over Kevin O’Leary’s planned data center. A nonprofit and five residents are challenging the constitutionality of the approval process for the Stratos Project. O’Leary has already agreed to scale back the plan.
Policy & Regulation
White House signs executive order on frontier AI innovation and security. The order pushes deployment, hardens cybersecurity, and sets up a voluntary federal review process for covered frontier models. The week’s most consequential primary document on US AI oversight. A useful companion analysis from Tech Policy Press unpacks the politics behind it.
Anthropic proposes giving governments power to block dangerous frontier AI deployments. Two new policy frameworks cover catastrophic risks (bio, cyber, loss of control) and the economics of AI, including how to prepare workers and spread the financial gains.
Bipartisan “Great American AI Act” draft would preempt state AI laws for three years. The Obernolte-Trahan discussion draft proposes a federal Center for AI Standards and Innovation plus mandatory third-party audits of frontier models. Legal and policy circles spent the week dissecting it.
New York becomes first state to require disclosure of AI performers in ads. The law mandates transparency when ads, film, or TV use synthetic actors, targeting consumer deception and the displacement of creative workers.
Colorado restricts AI therapy chatbots, joining a wave of state action. The new law requires psychotherapy to come from a licensed professional and limits AI to supporting roles. Chatbots also cannot market themselves as equivalent to therapists or imply HIPAA-style confidentiality.
Bruegel says the EU AI Act is regulating the wrong things. The brief argues the law’s upfront risk classifications miss real-world harms like chatbot-induced self-harm and election misinformation, and proposes lighter compliance burdens in exchange for stronger liability and a public AI incident registry.
State governments are deploying AI faster than they can govern it. US states are putting generative AI into public-sector work despite thin transparency rules, unresolved procurement questions, and weak data governance.
Copyright & The Courts
UK competition regulator forces Google to let publishers opt out of AI search. An early precedent for how regulators may handle AI’s extraction of publisher value, framing it as a market-power problem rather than just a copyright one. Tech Policy Press has a sharp analysis of why this matters for journalism’s economics.
Google argues YouTube’s terms of service let it train AI on musicians’ work. Its motion to dismiss the Lyria 3 class action claims artists granted a broad royalty-free license the moment they uploaded their tracks. A defense with implications for everyone who has ever agreed to a platform’s terms.
Musicians union sues UMG and Warner over unshared AI settlement money. The AFM says the major labels pocketed payouts and licensing deals from Udio and Suno while refusing to compensate the musicians whose recordings trained those systems.
Judge unseals Udio’s AI training data figures. The ruling could force public disclosure of exactly how many copyrighted recordings trained the music generator, a win for transparency in AI copyright litigation.
UK MP sues xAI over nonconsensual Grok deepfakes. Backed by more than 100 advocacy groups, Jess Asato’s case aims to set a precedent holding companies liable for the design and foreseeable harms of their AI systems.
Economics & Employment
Anthropic pledges $200 million to study AI’s economic fallout. The money funds research into job displacement and economic transformation, while the CEO floats policy ideas for labor markets. A sign the industry is taking the automation question seriously, or at least wants to look like it.
BCG: 74% of frontline workers now use AI regularly. The fourth annual survey of nearly 12,000 employees finds adoption up over 20 points in two years, with jobs changing faster than companies can redesign work. Frontline staff increasingly expect to supervise AI agents rather than do tasks themselves.
Deloitte names 82 white-collar jobs where AI is already slowing hiring. Software programming, database administration, and technical sales top the list, with entry-level workers taking the hardest hit over the next five years.
Brad Smith on AI, jobs, and the next generation. Microsoft’s president reports 17.8% global generative AI usage (31.3% in the US) and argues for preparing young workers to use the technology rather than be displaced by it.
Brookings warns the global AI divide could harden into permanent inequality. With infrastructure, data, and talent concentrated in the Global North, the Global South faces job displacement and skills mismatches affecting the 61% of global workers in informal employment. AI could still add $1.2 trillion to Africa’s economy by 2030 if policy gets it right.
Ethics & Safety
Bots now generate more web traffic than humans, Cloudflare reports. Automated requests hit 57.5% of HTTP traffic, a milestone arriving 18 months earlier than CEO Matthew Prince predicted. Since bots do not click ads, the web’s basic business model is now in question.
AI floods the courts with self-represented litigants. A Colorado magistrate judge reports self-represented filings have more than doubled since 2023, a jump she pins on chatbots. AI seems to widen access to the legal system without improving anyone’s odds of winning.
AI will not start a nuclear war, but humans using it might. Researchers argue the real near-term risk is not rogue AI launching weapons but scaled deepfakes and disinformation misleading leaders into escalation, plus hallucination-prone models in high-stakes military settings.
MIT: AI is making us worse at spotting fake news. New Media Lab research finds that leaning on AI erodes our ability to detect misinformation, much the way GPS weakened our sense of direction.
Research
Nature poll: scientists have a bad case of AI FOMO. Of 1,900+ researchers surveyed, nearly half feel broadly negative about AI and 63% say the risks of LLMs outweigh the benefits. Many adopt the tools anyway, out of fear of falling behind.
New automation index says data scientists are more exposed than plumbers. The preprint computes occupational automation scores showing non-routine cognitive roles at high risk while unstructured trades stay resilient, upending the standard job polarization story.
What happens when everyone asks the same AI for advice. This paper examines the societal risks of delegating health, legal, financial, and personal decisions to generative AI, including behavioral homogenization, eroded autonomy, and a shrinking diversity of judgment at scale.
Last Updated: 2026-06-12 07:29 (California Time)