Back to News
PolicyAI Understanding briefing

Pennsylvania tightens disclosure rules for AI data centers as secrecy draws scrutiny

Tom’s Guide reports that Pennsylvania has barred state agencies from signing nondisclosure agreements with data-center developers and removed AI data centers from a fast-track permitting program. The move follows growing concern that communities may not know who is behind projects or how much electricity, water and…

By 6 min read
AI-generated editorial illustration accompanying Pennsylvania tightens disclosure rules for AI data centers as secrecy draws scrutiny
The short version

Tom’s Guide reports that Pennsylvania has barred state agencies from signing nondisclosure agreements with data-center developers and removed AI data centers from a fast-track permitting program. The move follows growing concern that communities may not know who is behind projects or how much electricity, water and…

What happened

Tom’s Guide reported on August 24, 2026, that Pennsylvania Gov. Josh Shapiro signed an executive order on August 18 imposing new rules on AI data-center development. The order prevents state agencies under his authority from entering nondisclosure agreements with developers, removes AI data centers from fast-track permitting and calls for greater public disclosure and community involvement. Tom’s Guide also described similar secrecy concerns in Virginia, Texas and Wisconsin, while reporting that Microsoft has said it will stop using NDAs with local governments for data-center development.

Tom’s Guide published the report on August 24, 2026, identifying Pennsylvania’s August 18 executive order as the clearest concrete policy action in the article. According to Tom’s Guide, state agencies under Gov. Josh Shapiro’s authority can no longer enter nondisclosure agreements with data-center developers. The order also removes AI data centers from the state’s fast-track permitting program and requires more public disclosure and community involvement before projects proceed. These details are attributed to Tom’s Guide and are not independently confirmed here through the executive order itself.

Tom’s Guide frames the Pennsylvania action as a response to a broader pattern in which data-center developers have used NDAs, shell companies and code names during negotiations with local governments. The report says public officials in some cases signed agreements that prevented them from discussing projects that could eventually consume substantial amounts of community electricity. In Texas, Tom’s Guide says shell companies and NDAs have made it difficult for residents and sometimes state officials to identify the parties behind proposed projects. The article describes similar questions in Wisconsin, but does not provide a complete list of projects or agreements.

The report cites research from University of Mary Washington researchers who found that 25 of 31 Virginia localities with an existing, approved or proposed data center had entered nondisclosure agreements. Tom’s Guide says the researchers warned that the true number could be higher. The article does not identify all of the localities, describe the terms of each agreement or establish whether every agreement concerned AI-specific facilities rather than data centers more generally.

Tom’s Guide also reports that Microsoft said this year it would stop using nondisclosure agreements with local governments when developing data centers. The company reportedly said communities should know about its plans and acknowledged that greater transparency was needed to build trust. The article places that decision alongside Microsoft’s continuing investment in AI infrastructure, but it does not provide the terms of the company’s policy, identify every jurisdiction covered or independently confirm the announcement.

Read the primary source: tomsguide.com

Why it matters

AI infrastructure is becoming a local policy issue because data centers can affect electricity, water, land use, taxation and public construction. Tom’s Guide cites estimates that a modern AI data center can use as much electricity as 100,000 homes and that large facilities can consume millions of gallons of water daily, although the article does not independently verify those figures or provide project-specific measurements. Earlier public disclosure could give residents more information before major infrastructure and incentive decisions are made.

Tom’s Guide’s central point is that AI is not only a software or model issue. The systems behind services such as ChatGPT, Gemini and Claude require data centers filled with servers and GPUs, and those facilities require land, electricity and, depending on their cooling systems, water. When the identity, size or resource requirements of a proposed facility are hidden during negotiations, residents may have limited information about a project before public decisions are made.

The article cites the World Resources Institute as estimating that a modern AI data center can use as much electricity as 100,000 homes, while the largest facilities may use substantially more. Tom’s Guide also says large data centers can use as much as 5 million gallons of water per day. Those are attributed estimates, not measurements from a specific Pennsylvania, Virginia, Texas or Wisconsin project. The source does not independently confirm them, explain the range of facility designs involved or distinguish routine data-center demand from demand specifically caused by AI workloads.

The infrastructure consequences described by Tom’s Guide extend beyond the buildings themselves. A large facility may require substations, transmission lines, roads and water infrastructure, and local or state governments may offer tax incentives to attract development. If residents do not know a project’s expected size and resource demands, they may be unable to assess whether projected investment and jobs justify long-term costs. The article does not provide a cost-benefit analysis of any specific project or establish who would ultimately pay for each type of infrastructure.

The Pennsylvania order therefore matters as a transparency measure rather than as evidence that AI data centers have been stopped. Tom’s Guide reports that Pennsylvania had been actively courting data-center investment, including Amazon’s previously announced plan to invest at least $20 billion in data-center campuses in the state. The source does not say whether that plan is affected by the order, whether any project has been cancelled or whether the new rules will materially slow construction. The practical effect remains an important unknown.

What to watch next

The key questions are how Pennsylvania’s rules will be implemented, whether other states adopt comparable limits on secrecy, and whether local governments receive specific information about proposed facilities. Tom’s Guide says residents may need to examine planning agendas, land sales, tax agreements, utility filings and water or air permits because projects can appear under shell companies or code names. The source does not establish how many projects will be affected, whether local NDAs will be prohibited in every jurisdiction, or whether transparency requirements will change project approvals.

Implementation will be the first test. Tom’s Guide reports what the Pennsylvania order changes for state agencies, but it does not explain how agencies will define an AI data center, what disclosures developers must provide, how quickly information must be released or what enforcement follows if a project is not transparent. It also does not establish whether the order applies to local governments or only to agencies under the governor’s authority.

Other states may face pressure to adopt similar policies, but the article documents concern rather than a coordinated national movement. Tom’s Guide identifies secrecy questions in Virginia, Texas and Wisconsin and describes Pennsylvania’s action as part of a growing backlash. It does not report additional state laws, votes or executive orders that would demonstrate a broader legal trend. Whether other states act will depend on local permitting rules, utility constraints, tax incentives and community opposition, none of which are fully detailed in the source.

For residents, Tom’s Guide recommends looking beyond conventional company announcements. The report says proposed facilities may first appear in planning and zoning agendas, land-sale records, tax-abatement agreements, utility filings or applications for water and air permits. Projects may be listed under unfamiliar limited-liability companies or code names such as “Project Nova,” meaning searches for major technology companies alone may miss relevant records. The source does not provide a jurisdiction-specific public-records guide or confirm that every project using a code name is an AI data center.

The most consequential facts to obtain are the facility’s expected size, electricity and water demand, the identity of its eventual operator, the timing of public review and responsibility for new substations, transmission lines, roads and water systems. Tom’s Guide says residents can ask what happens to those costs if a project is delayed, downsized or abandoned. The article does not report answers from Pennsylvania officials, developers or utilities, so the effects of the order and Microsoft’s stated policy remain untested in the material provided.

Related guides & quizzes

AI EthicsFuture of AIAI TrainingTest what you know — try a free AI quizLook up an AI term in our glossary
Found this useful?