A campaign pledge joins two political debates

The AI Pact is attempting to turn a diffuse backlash against data-centre development into a coherent electoral platform on artificial intelligence. As of August 26, 2026, the initiative listed 18 candidates for office as signatories, including Nebraska Senate candidate Dan Osborn and Texas gubernatorial candidate Gina Hinojosa. Its central proposition is that the physical expansion of AI infrastructure and the governance of AI systems should be addressed together.

That is a notable framing. Arguments over data centres have often been conducted through the language of zoning, tax incentives, water consumption, transmission lines and electricity bills. Debates over AI, meanwhile, have focused on safety testing, misinformation, discrimination, employment and legal liability. The pact treats these as connected consequences of the same industry expansion: communities host the facilities, while workers and consumers bear risks associated with the technologies those facilities enable.

The approach gives candidates a way to speak about AI in terms that are tangible to voters. A proposed data centre is visible, resource-intensive and locally governed; a frontier model and its potential effects may appear remote. Bringing the two together may therefore broaden the political constituency for AI rules, especially in areas facing fast-growing demand for power infrastructure.

Five promises, ranging from local development to national rules

The pact asks candidates to endorse five broad commitments. Its data-centre provision opposes tax breaks, opaque agreements and excessive energy use. The other four call for independent safety reviews before the release of major new AI models, an “AI dividend” for workers displaced by automation, legal rights to sue AI companies for material or financial harm, and opposition to public bailouts for AI firms.

The platform is deliberately more of a statement of principle than a legislative blueprint. It does not define which models qualify as “major”, set a testing methodology, specify how an AI dividend would be funded, or establish the legal threshold for harm. Those gaps make it easier for candidates from different districts and levels of government to sign, but they will become important if supporters seek to translate the pledge into law.

The data-centre promise is similarly a floor rather than a blanket ban. It does not require a nationwide moratorium. That distinction matters politically: construction projects can bring short-term employment, new tax revenue and investment in areas seeking economic development. Yet the facilities can also require major upgrades to generation and transmission, consume substantial volumes of water depending on their cooling design, and raise concerns over noise, land use and the allocation of utility costs.

Power costs have become the immediate pressure point

The energy dimension has made this issue unusually salient. A Department of Energy-backed assessment estimated that data centres used about 4.4% of US electricity in 2023 and could account for 6.7% to 12% by 2028. The wide range illustrates both the scale of projected growth and the uncertainty surrounding the pace of AI deployment, equipment efficiency and new construction.

Federal and state policymakers are already responding, though not necessarily through the pact’s preferred approach. In June, the Federal Energy Regulatory Commission directed regional grid operators to justify or reform rules for connecting large electricity users, including data centres. The agency’s focus includes cost-allocation arrangements intended to avoid leaving residential customers responsible for infrastructure built for a project that is delayed or never materialises.

The White House has also promoted a voluntary Ratepayer Protection Pledge under which data-centre developers, utilities and large technology companies commit to fund the new generation and delivery infrastructure required for their projects. Supporters see such arrangements as a practical way to preserve investment while protecting household bills. Critics of voluntary pledges, including advocates of stronger regulation, argue that public commitments do not substitute for enforceable disclosure requirements, permitting standards or liability rules.

Texas shows how quickly the politics have shifted. Governor Greg Abbott directed state regulators and grid operator ERCOT to audit prospective data-centre projects and pause approvals while information is reviewed. The state’s requirements focus on projected electricity demand, plans to reduce grid reliance, water use and reuse, community impacts, ownership disclosure and public incentives. It is not a general prohibition on data centres, but it reflects a growing consensus that large projects must show how they will pay for their effects.

The difficult link between AI safety and infrastructure

The pact’s most ambitious move is to connect local infrastructure safeguards with national AI safety policy. The connection is intuitively compelling: computational capacity is a prerequisite for training and operating powerful systems, and the rapid build-out of facilities can be seen as evidence of a race to deploy them. But the policy tools operate at different levels.

A city or state can influence land use, water permits, tax abatements and utility terms. A meaningful pre-release review of advanced AI models would generally require national standards, qualified independent evaluators, secure access to models and clarity about what failures trigger remediation or delay. Likewise, a private right of action could create accountability, but its practical impact would depend on the definition of causation, available evidence and limits on liability.

This does not make the combined agenda incoherent. It instead exposes the need for coordination. Communities need accurate information before committing land, water and grid capacity; regulators need credible estimates of future load; and the public needs clarity about the social and economic value expected in return. Candidates who sign the pact are making an explicit choice to judge AI infrastructure not only by its investment totals, but also by its distributional consequences.

An early test of electoral durability

With only 18 signatories listed, the AI Pact remains a small campaign initiative rather than a major cross-party coalition. Its current profile is also uneven: most signers are Democrats, while Osborn is an independent. That limits claims of broad partisan alignment, even though concern about electricity prices, local control and corporate subsidies can resonate across ideological lines.

Its significance lies less in the number of signatures than in the emerging vocabulary. Data-centre politics is no longer solely about attracting investment or speeding permits. It increasingly includes questions over who pays for power, whether local residents receive meaningful consultation, how workers share in productivity gains, and whether AI developers can be held accountable for harms.

The pact offers candidates a concise answer to those questions. Its next challenge is more demanding: converting a set of popular commitments into specific rules that can withstand scrutiny from utilities, local governments, labour groups, technology companies and the voters whose communities will host the infrastructure.

Sources