OpenAI Drops NDAs, Pledges Transparency on Data Centers After Midterm Backlash

OpenAI published a pair of policy documents on Thursday committing to greater transparency and community control over its future data center projects, two days after voters across 10 states rejected or restricted new AI infrastructure in nearly every local ballot measure.

The company released a National Community Compact — voluntary promises for campuses it owns and operates — and a Compute Campus Policy Blueprint urging lawmakers to turn similar rules into binding law. The announcements follow midterm elections in which Democrats flipped governorships in Texas and Georgia, won control of both congressional chambers, and saw voters treat data centers as a decisive issue.

Election results force the issue

According to Ballotpedia analysis and CNBC reporting, 41 local measures across 10 states went before voters on Tuesday. All but one — Harwood, North Dakota — backed restrictions or rejected pro-development measures. In Ohio alone, 21 measures appeared on ballots. Democrat Amy Acton won the governor's race after supporting a "conditional moratorium" on data centers until they cover 100% of utility costs and "exclusively build with union labor." In Texas, Gina Hinojosa unseated Gov. Greg Abbott after making data center opposition central to her campaign [3].

A CNBC Voter Poll found 56% of voters called data centers a crucial or major factor in their decisions, with Democrats ranking the issue higher than Republicans [3]. The results prompted immediate calls from House Democrats for AI executives to testify under oath [3].

What OpenAI actually committed to

The National Community Compact applies to "every campus owned and operated by OpenAI" [11], and OpenAI's blog post specifies "future campuses that we own and operate" [7]. It does not cover sites where OpenAI is a tenant or partner. The commitments, in OpenAI's own words [11][7]:

• No NDAs with public officials, allowing only "narrow protections covering legitimate security information, trade secrets, or proprietary or competitively sensitive information."
• Pay full electricity and infrastructure costs, with financial assurances posted before major spending begins so ratepayers are protected if a project is delayed or canceled.
• Publish per-site water and electricity use, construction plans, and independent noise studies before operations begin and annually afterward.
• Prioritize closed-loop or low-water cooling and develop community-specific water plans.
• Solicit community input through surveys, listening sessions, and a designated local point of contact.
• Publish workforce strategies and annual hiring reports, work with unions and training providers, and invest in community-selected priorities (childcare, education, healthcare, transportation).

The Compute Campus Policy Blueprint asks federal, state, and local policymakers to codify similar rules industry-wide: barring NDAs for public officials, mandating public reporting on water, energy, noise, and economic impacts, protecting ratepayers through large-load tariffs, and requiring transparent tax incentive agreements with clawback provisions [12].

Scope limits and open questions

Several limitations are worth noting. The compact covers only future OpenAI-owned and operated campuses — not existing projects, leased capacity, or partner-run sites like the Ohio campus developed with SoftBank and SB Energy [3]. The ownership status of the company's Saline, Michigan project with Oracle and its Abilene, Texas site under the new framework is not explicitly addressed in the documents released Thursday [9][11][12].

OpenAI says the blueprint has been in development since August and builds on community commitments announced in January and site-specific pledges in Georgia and Ohio [7]. Chris Lehane, OpenAI's Chief Global Affairs Officer, told reporters the documents reflect "the best lessons that we have learned and committing to applying them as they go forward" [1].

The NDA pledge includes a carve-out for proprietary information. Where the line falls between protected business details and public-interest data (water demand, power contracts) will determine how much transparency actually improves [4]. The compact also lacks an enforcement mechanism beyond reputational pressure; compliance will be self-reported annually.

Industry context

Amazon announced in October it would stop using NDAs with government agencies on data center projects [5]. Meta, Microsoft, Google, and Anthropic have outlined similar voluntary commitments [1]. OpenAI's blueprint explicitly calls for industry-wide rules — a move that levels the playing field if competitors adopt them, but leaves OpenAI exposed if they don't.

What remains unknown

• How many future campuses OpenAI will own versus lease, and thus how much of its buildout the compact covers.
• Whether the first local compacts will include measurable targets, independent verification, and named contacts — or read like boilerplate.
• Whether state utility commissions will adopt the blueprint's large-load tariff proposals, and on what timeline.
• How the confidentiality exception for "competitively sensitive information" will be interpreted in practice.

The midterm results handed AI companies a clear political signal: communities want veto power, not just transparency. OpenAI's response opens the door to that conversation. Whether it delivers substance or symbolism will be visible in the first local compacts and the first annual reports.