A.I. Meets Resistance on Power, Labor and Oversight

On a single day, the fast-expanding artificial intelligence industry ran into resistance on three fronts that increasingly define the next phase of the technology’s rise: the land and power it consumes, the workers it may displace or rank, and the question of who, if anyone, should be empowered to slow it down.

In New York, Gov. Kathy Hochul imposed a one-year moratorium on new “hyperscale” A.I. data centers, pausing environmental permits for facilities requiring 50 megawatts or more as the state drafts rules to address their effects on the electric grid, water supplies, the environment and utility customers. In California, 26 current and former Meta employees sued the company in federal court, arguing that A.I.-assisted layoff and productivity systems disproportionately harmed workers with disabilities or those on protected leave. And in Washington, Google DeepMind’s chief executive, Demis Hassabis, urged the United States to create a national A.I. standards body or watchdog to evaluate the most powerful models before release.

The developments, though distinct, point to a broader shift. The A.I. debate is no longer centered mainly on speculative arguments over whether regulators will stifle innovation. It is increasingly being fought over practical questions: how much electricity and water the industry should be allowed to use, whether algorithmic tools can be trusted in workplace decisions, and whether voluntary promises from technology companies are enough as systems grow more capable.

New York Tests the Limits of A.I. Buildout

New York’s move is among the clearest signs yet that state governments may begin to challenge the physical expansion required to support the A.I. boom.

The state’s temporary ban applies to new facilities classified as hyperscale, a category aimed at the largest data centers that underpin training and deployment of advanced models. Officials have said the pause will give regulators time to study whether such projects could strain the grid, increase costs for ratepayers and intensify pressure on local water resources.

Those concerns have been building as utilities across the country confront a surge of demand from data centers, electric vehicles and manufacturing. In New York, more than 12 gigawatts of very large power loads were waiting to connect to the grid as of May, according to earlier reporting by Reuters — a figure that has sharpened anxiety over whether the state can accommodate a new class of energy-hungry computing campuses without broader consequences.

The moratorium lasts one year, and its long-term significance remains uncertain. It could become a template for other states or local governments seeking leverage over the A.I. industry’s infrastructure footprint. Or it could prove a temporary political warning shot in a state trying to buy time before approving a new wave of projects.

Either way, the action reflects a widening realization that A.I. is not only software. It is also warehouses of servers, transmission constraints, water-intensive cooling systems and land-use fights — all of which place the industry into more familiar regulatory terrain.

A Lawsuit Over A.I. and Layoffs

The lawsuit against Meta raises a different but increasingly urgent question: when companies use A.I.-driven tools to rank workers or shape personnel decisions, who bears responsibility if those systems produce discriminatory outcomes?

The plaintiffs, 26 current and former employees, allege that Meta used A.I.-assisted layoff tools and productivity metrics in ways that disproportionately targeted people with disabilities or workers who had taken protected leave. The case was filed in federal court in Oakland. Meta has denied that A.I. made the decisions, saying that workforce choices “were and are made by people, not AI.”

That defense may become a central fault line not only in the case, but in the broader legal treatment of algorithmic management. If a company relies on machine-generated scores, risk flags or performance rankings, courts may eventually have to decide whether those systems are simply inputs into human judgment or whether they represent a distinct mechanism of decision-making that can trigger discrimination liability on their own.

The case arrives as labor economists and researchers are increasingly warning that the effects of A.I. on work are moving from theory to lived experience. This week, more than 200 researchers and economists called for urgent policies to address the technology’s economic and labor-market consequences, adding to pressure on lawmakers to look beyond model safety and into workplace disruption.

That shift matters because public debate over A.I. has often focused on existential risk, disinformation and national competition. But for many workers, the first encounter with the technology may come not through a chatbot, but through scheduling systems, productivity monitoring or software used to decide who stays and who goes.

Even A.I. Leaders Are Asking for Rules

Mr. Hassabis’s call for a U.S.-led standards body underscored another emerging reality: some of the industry’s most prominent figures are now arguing that stronger oversight is not just inevitable, but necessary.

He said urgent action was needed as A.I. capabilities accelerated, and suggested that a national body could evaluate frontier models before they are released. Such an institution, in his view, could also help coordinate slowdowns if systems begin to pose heightened risks.

The proposal goes beyond the federal government’s current approach, which has largely emphasized voluntary commitments and informal discussions with major A.I. companies. Reuters reported earlier this month that U.S. officials were in talks with companies about voluntary model standards, a path that reflects both the political difficulty of creating a new regulator and the government’s desire to avoid constraining a strategically important industry too quickly.

But Mr. Hassabis’s intervention suggests that the center of gravity may be moving. When a leading executive at one of the world’s foremost A.I. labs begins advocating a formal standards-setting or watchdog function, it signals that arguments once dismissed as anti-innovation are gaining credibility even inside the sector.

Whether Washington is prepared to act on that is another matter. The United States has often preferred light-touch governance in emerging technologies, and lawmakers remain divided over how aggressive federal intervention should be. A stronger oversight body would also raise difficult questions about independence, enforcement power and how to balance safety with competition against China and other rivals.

A New Phase of the Backlash

Taken together, the three developments show an industry entering a more complicated political era.

The first phase of the A.I. boom was driven by breakthroughs in generative systems and the race among technology giants to build bigger models faster. The next phase appears likely to be shaped by a different set of actors: utility regulators, judges, state officials, labor advocates and, increasingly, executives who want clear rules before public trust erodes further.

What makes this moment notable is not simply that criticism of A.I. is growing. It is that the criticism is becoming operational. It is attaching itself to permits, court filings, grid queues, disability law and institutional design.

For the industry, that may prove harder to brush aside than generalized concern. A temporary state moratorium can delay infrastructure. A discrimination suit can force disclosure of how internal systems actually work. And calls for a federal standards body can give political cover to policymakers who had been reluctant to move beyond voluntary guardrails.

The result is a more grounded backlash — one less about abstract fear of machine intelligence and more about who pays, who decides and who is protected as A.I. spreads through the economy.