Candidates Sign AI Pact Targeting Data Centers and Safety

More than 15 political candidates across the United States have signed the AI Pact, a new pledge containing five core promises aimed at regulating artificial intelligence and the massive data centers that power it. Organized as an offshoot of the Political Integrity Project, founded by Daniel Lobo-Lewis, the initiative is funded by a progressive political action committee rather than tech companies or AI safety groups. Signatories include Nebraska independent Senate candidate Dan Osborn, Tennessee Representative Justin Pearson, and Michigan candidate Will Lawrence.

The pact's five promises require candidates to advocate for mandatory safety reviews of AI models, support policies establishing an AI dividend for workers, and allow consumers and companies to sue over AI-related harms. On the infrastructure side, the pledge opposes tax breaks, backroom deals, and excessive energy usage for data centers. This political push reflects rising bipartisan concern over the environmental and economic impacts of AI infrastructure. For instance, Texas Governor Greg Abbott recently initiated a temporary moratorium on data center construction, while Texas Attorney General Ken Paxton proposed making data centers legally liable for harmful chatbots.

For AI practitioners, developers, and data center operators, this growing political movement signals a shift from voluntary guidelines to hard regulatory compliance. While the White House has focused on voluntary safety frameworks and OpenAI CEO Sam Altman has discussed worker dividends with political leaders, the AI Pact advocates for legally binding mandates. If these candidates succeed, tech companies could face strict environmental limits on energy consumption, the loss of lucrative local tax incentives, and direct legal liability for model outputs. Developers may soon have to factor mandatory, state-level safety audits and potential litigation risks into their deployment timelines, ending the era of rapid, unchecked scaling.