AI giants and unions launch an alliance to write data center rules
Three stories this hour about who gets a say before the buildout lands: a coalition trying to write the rulebook itself, a government tightening the leash on the people who write the models, and a court that finally put a price on the compute an AI consumes.
Blackstone, OpenAI, QTS and SoftBank are backing the American Infrastructure Alliance, a new labor-and-business coalition that wants to write data center standards with state legislators before the 2027 sessions begin. Axios reported the launch on Monday and the coalition's own release confirms it: the group describes itself as the first nationwide coalition of labor and business on this issue, initial campaigns are underway in seven states — Texas, Georgia, Ohio, Iowa, Pennsylvania, Indiana and South Carolina — and it says its work "will begin ahead of and continue through the 2027 state legislative sessions," explicitly not a midterm campaign. Members include the electrical workers' IBEW, the Iron Workers, the plumbers' UA and the sheet metal workers' SMART, alongside QTS, SoftBank, SB Energy, CoreWeave, Prologis, Related Digital and Digital Realty.
The pitch is that voluntary standards are cheaper than referendums. Its own polling found Ohio voters rate data centers 18 percent favorable against 67 percent unfavorable, but flip to roughly 60–34 in favor when guardrails and local benefits are attached — the same shift we reported in America has turned against data centers — Pew finds 54% now call them bad for the environment. The coalition says it will publish principles in the coming weeks: protect water and local resources, pay your own way, good jobs, visible local benefits, operator accountability. QTS co-CEO Tag Greason frames it as "clear expectations for everyone involved"; IBEW international president Kenneth Cooper argues that "blanket bans on necessary infrastructure projects would set back our economy and threaten good middle-class jobs." Both sides know what they are buying. Nothing here is binding — it is a promise to negotiate, priced at a fraction of what a state moratorium costs, and the industry has been on the losing end of those votes for a year now. Two things to watch: whether the alliance puts a dollar figure behind "benefits," and whether the communities already fighting projects see a seat at the table or a rival they now have to negotiate past. We tracked the money version of this backlash in local opposition blocked $68B of US data centers in one quarter.
China has widened its overseas travel restrictions to cover the direct relatives — spouses and children — of certain AI and chip executives, requiring Beijing's approval before they travel abroad even for short trips. Bloomberg reported the change, citing people familiar with the matter: government agencies have recently begun notifying affected people, including prominent startup founders and heads of strategically important AI companies, and the plan is to add individuals to the list over time. It is a pre-approval requirement rather than an outright ban, and the report says it is unclear whether every affected person's family is covered. The legal hook is the exit-entry administration regulation that took effect on September 15, which allows exit bans tied to potential breaches of industrial and technological security, and the Ministry of Industry and Information Technology did not respond to a request for comment. This is a single-source Bloomberg exclusive for now — no company or executive has been named, and no second outlet has independently confirmed the family provision — so treat the specifics as reported, not settled.
A district court in Wuhan counted AI production costs — token consumption plus commercial AI-tool licensing fees — in a copyright damages calculation, which Chinese media describe as the first time that has happened. Per a National Law Review analysis of the ruling, the Jiang'an District People's Court decided the case on August 20 and announced it in late September: a producer made a 47-episode AI-assisted short drama in early 2026, and a day after release another company copied the whole series, retitled it and ran ads against it. The court held the drama was a protectable audiovisual work because staff made creative decisions at every stage — script, storyboard prompts, selecting AI outputs, editing and subtitle syncing — leaving the AI as a means rather than an author.
The damages number is small: 20,000 RMB, about $2,900. The reasoning is the part worth keeping. Neither side proved actual losses or the infringer's profit, so the court set discretionary damages weighing runtime, distribution reach, release window, how long the infringement lasted and fault — and when it assessed production cost, it included the compute bill from tokens and tool licences. That is the first time a court has treated AI spend as a cost of making the thing being stolen, and the court's own advice to creators is the practical takeaway: keep the scripts, the prompt drafts, the generation records and the project files, because that is now the evidence that decides authorship.
What to watch: whether the alliance publishes its principles with numbers attached, and whether a second outlet corroborates the family travel restrictions — Beijing has not commented.
Should AI companies be allowed to write data center standards before they get broken by referendum, or is that the fox drafting the henhouse rules? Tell us in the comments.
Sources: Axios · American Infrastructure Alliance · Bloomberg · Financial Times · National Law Review · The Decoder · Changjiang Daily (长江日报)
