According to a new disclosure filed Tuesday, OpenAI spent $1.76million on government lobbying and $510,000 during the last three months alone of the year. This is a significant increase from 2023 when the company only spent $260,000 on Capitol Hill. OpenAI also revealed a new lobbyist, Meghan Dorn, who spent five years working for Senator Lindsey Graham before joining the company in October. The filing also shows activities related to two new pieces legislation in the last months of the year. The House’s AI Advancement and Reliability Act would establish a government center to conduct AI research. The Senate’s Future of Artificial Intelligence Innovation Act would create benchmark tests for AI models. OpenAI did not answer questions about its lobbying activities.
Perhaps more importantly, the disclosure is a signal that the company has arrived as a player in politics, as its first serious lobbying year ends and Republican control begins in Washington. OpenAI’s lobbying expenditures are still dwarfed compared to its peers–Meta is the top Big Tech spender, with over $24 million by 2024- but the increase comes as the company and other AI companies helped redraw AI policy.
AI policy has resembled a whack a mole response to the risks of deepfakes, misinformation and other fake news for the past few decades. Over the past year, AI companies began to argue that the success of this technology is crucial to American competitiveness and national security, and that the government should therefore support its growth. OpenAI, among others, is now poised to access cheaper energy, lucrative contracts for national security, and a regulatory environment that doesn’t care about the details of AI safety.
Although the major players appear to be more or less in agreement on this grand narrative the messy divisions on other issues still threaten to break the harmony displayed at President Trump’s Inauguration this week.
AI regulations began to take shape after ChatGPT was launched in November 2022. Liana Keesing is the campaigns manager for Issue One’s technology reform, a democracy non-profit that tracks Big Tech influence. She says that at that time, “a lot was said about responsibility.”
Companies are asked what they will do about sexually abusing deepfake pictures and election disinformation. Keesing says that Sam Altman did an excellent job of portraying himself as a proponent of the process.
OpenAI began its official lobbying efforts around October 2023. Chan Park, a former Senate Judiciary Committee Counsel and Microsoft lobbyist, was hired to lead the effort. According to OpenSecrets, lawmakers, notably then Senate majority leader Chuck Schumer were vocal about their desire to curb these specific harms. OpenAI hired Schumer’s former legal counselor, Reginald Babin as a lobbyist. The company hired Chris Lehane, a veteran political operative as its head for global policy this past summer. OpenAI’s disclosures from the past confirm that OpenAI’s lobbyists spent a large part of the last year working on legislation such as the No Fakes Act or the Protect Elections from Deceptive AI Act. The bills never became law. As the year progressed, the regulatory goals for AI companies began changing. Keesing says that “one of the biggest changes that we’ve noticed” is that they’ve started to focus on the energy.
Altman, with leaders from Nvidia and Anthropic and Google, visited White House in September and pitched the vision, that US competitiveness will depend on subsidized infrastructure for training the best AI models. Altman proposed the Biden administration to build multiple five-gigawatts data centers that would consume as much energy as New York City.
Around that time, companies such as Microsoft and Meta began to claim that nuclear energy would be the future of AI. They announced deals aimed at launching new nuclear power stations.
OpenAI’s policy group was likely already planning for this shift. Matthew Rimkunas was hired by the company in April. He previously worked for Bill Gates’ sustainable energy initiative Breakthrough Energies. Before that, he spent 16 years as a lobbyist for Senator Graham, a South Carolina Republican who serves on the Senate Subcommittee for Nuclear Safety.
The new AI energy race cannot be separated from the positioning AI as essential to national security and US competition with China. OpenAI outlined its position in an October blog post, writing that “AI is a transformative technology that can either be used to strengthen or undermine democratic values.” In December, OpenAI announced that it would work with Anduril, a defense-tech company, to develop AI models to help destroy drones near military bases. In the same month, Sam Altman stated in an interview with The Free Press, the Biden administration had been “not very effective” at fostering AI. “The things I think should have the administration’s priority, and I hope the next administration will be the priorities, are building massive AI infrastructure in America, having a US supply chain, things like that.” It also preceded a recent executive order issued by Biden to lease federal land for the type of data centers Altman was asking for. Altman’s posture, whether it was on purpose or not aligned him to the growing camaraderie that President Trump has with Silicon Valley. Altman, Elon Musk and Jeff Bezos sat behind Trump’s family on Monday at the inauguration. Mark Zuckerberg and Elon Musk were also in attendance. Altman, who personally donated $1 million to Trump’s inaugural funds, was one of many tech leaders who had made substantial donations.
The inauguration is a good way to see that these tech leaders have aligned themselves with each other and other players in Trump’s orbit. There are some key dividing points that will be interesting to watch. There’s a clash over H-1B Visas, which allows many noncitizen AI Researchers to work in the US. Musk and Vivek RAMASWAMY (who, as of this past week, is no longer part of the Department of Government Efficiency), have been pushing to expand that visa program. Steve Bannon was the most vocal of Trump’s allies to criticize this.
The battle between open-source and closed-source AI is another fault line. Google and OpenAI don’t want anyone to know what their most powerful models are. They claim that this prevents them from being misused by bad actors. Musk has sued OpenAI over the issue. He claims that closed-source models are incompatible with OpenAI’s hybrid nonprofit organization. Meta, whose Llama open-source model recently sided with Muskin that lawsuit. Marc Andreessen, a Trump ally and venture capitalist, echoed these criticisms of OpenAI just hours after Trump’s inauguration. Andreessen also said making AI models open source “makes overbearing regulation unnecessary.”
Lastly, there are battles over bias. The different approaches taken by social media companies to moderate content, including Meta’s recent announcement to end its US fact checking program, raises questions about the future of AI models. Musk has lamented the “wokeness,” of many leading models. Andreessen said Tuesday that “Chinese LLMs were much less censored” than American LLMs (though this isn’t quite true, since many Chinese AI models are subject to government-mandated censorship that prohibits certain topics). Altman was more ambiguous: “No one will ever agree that a system is completely unbiased,” he said to The Free Press. The White House has been very busy, even though it’s just the beginning of a new Washington era. The White House has revoked many executive orders that President Biden signed, including the landmark AI order that imposed rules on government use of this technology (while appearing to have kept Biden’s order regarding leasing land for additional data centers). Altman is also busy. Altman was present when President Trump announced the project. Altman, according to Axios, will also be a part of a closed door briefing on January 30 with government officials. The meeting is reportedly about OpenAI’s development of an AI agent.
