Controlling the AI Narrative: The New Media Monopoly

AI companies have significantly increased their federal lobbying over the past two years, though some circulated growth figures overstate the pace.

The number of companies lobbying on AI issues grew from 458 in 2023 to 648 in 2024 — an increase of roughly 41% (some reporting on this figure mislabeled it a "141% increase," which the underlying numbers don't support) ^[7]. OpenAI's own lobbying spend rose nearly sevenfold, from $260,000 in 2023 to $1.76 million in 2024 ^[8] — still well behind Meta's $24 million-plus lobbying budget that year ^[7].

Nvidia's Q3 2025 in-house lobbying spend was $1.9 million, roughly 55% of its full-year 2025 lobbying budget of $3.46 million, coinciding with fights over AI chip export controls to China ^[9]. (An earlier claim that this figure "tripled" a specific Q2 2025 total could not be independently verified and has been removed.)

AI companies are also hiring former congressional staffers and regulatory officials to shape policy at every level of government. A recent academic analysis of AI industry influence tactics catalogued 27 distinct "capture mechanisms" across a sample of news coverage — including lobbying, revolving-door hiring, and framing regulation as anti-innovation or anti-competitiveness — finding at least 10 high-profile revolving-door cases in the U.S., UK and EU combined ^[10].

Beyond lobbying, AI executives and companies have made real and substantial political contributions — though several widely repeated dollar figures conflate different totals.

Trump's January 2025 inaugural committee raised more than $245 million overall — nearly 2.8 times his $88 million 2017 total. But that $245 million was the entire fund, not AI-specific: the Brennan Center found "technology interests gave roughly $44.6 million" of it, including $1 million each from OpenAI CEO Sam Altman, C3 AI, Perplexity AI, and Google ^[11].

The next day, Altman stood in the White House Roosevelt Room alongside Oracle's Larry Ellison and SoftBank's Masayoshi Son to announce the Stargate project, a planned $500 billion AI infrastructure initiative ^[12].

AI companies and their executives have built dueling super PACs aimed at the 2026 midterms, though the framing of a clean "regulation vs. innovation" fight doesn't hold up.

Anthropic donated $20 million to Public First Action in February 2026, later doubling that to $40 million. But contrary to some characterizations, Anthropic's own announcement states the money funds "public education and policy" work and explicitly cannot be used to support or oppose any candidate ^[13][14]. It was not framed by Anthropic as a counter to OpenAI's spending, and it does not fund a slate of 30–50 candidates — that specific claim does not appear in Anthropic's own materials or in contemporaneous reporting and has been removed.

The OpenAI-linked "Leading the Future" PAC, backed by OpenAI co-founder Greg Brockman and venture capitalists Marc Andreessen and Ben Horowitz ($12.5 million each), raised $50.3 million in the second half of 2025 and entered 2026 with $70 million in cash on hand ^[15][16] — a smaller sum than some reports have implied. Its stated goal is a single national AI regulatory framework that preempts state law, a position critics say functions as de facto deregulation even though it is not identical to opposing all AI rules outright.

The real dividing line isn't "pro-regulation" versus "pro-innovation" — it's federal preemption of state rules versus a patchwork of state laws, and neither side's PAC spending is remotely apples-to-apples with the other's.

The Trump administration's U.S. Tech Force program, run through the Office of Personnel Management under Director Scott Kupor, aims to recruit roughly 1,000 early-career engineers, data scientists and AI specialists into two-year federal placements, with salaries around $130,000–$200,000 ^[17].

More than two dozen tech firms are named as program partners, including Amazon, Apple, Google, Microsoft, Nvidia, OpenAI, Oracle, Palantir, Salesforce, Meta, Adobe and xAI ^[17] — the same companies building the AI systems the government is beginning to deploy and regulate.

Independent research on AI regulatory capture — rather than a single anonymous quote — documents the structural pattern: a 2026 academic analysis of industry-influence tactics found capture-related mechanisms (lobbying, revolving-door hires, narrative framing against regulation) present in the large majority of a sample of AI policy news coverage, with "regulation stifles innovation" and "national interest" among the most common industry arguments against binding rules ^[10].

Sam Altman has taken on a senior informal advisory role to the White House on AI, and stood alongside Larry Ellison and Masayoshi Son at the Stargate announcement the day after Trump's second inauguration ^[12].

OpenAI's embedding in federal infrastructure — Stargate, White House access, Tech Force partnership — mirrors a strategy other regulated industries have used: get close enough to government that separating regulator from regulated becomes politically difficult. That dynamic is well documented in the regulatory-capture literature even where individual quotes attributed to it in earlier drafts of this piece could not be verified and were removed.

On January 21, 2025, the day after Trump's inauguration, the President appeared at the White House to promote what was billed as a $500 billion AI infrastructure joint venture between OpenAI, SoftBank, Oracle and MGX, an Abu Dhabi sovereign fund ^[12].

SoftBank committed an initial $100 billion, with total commitments expected to reach $500 billion over roughly four years; the flagship Abilene, Texas campus alone is designed for 1.2 gigawatts, with more than 5 gigawatts of total capacity planned across U.S. sites, and the buildout is projected to create over 100,000 construction and operations jobs ^[12]. The $500 billion headline figure is in the same range as the annual GDP of a mid-sized economy such as Ireland (about $529 billion) — a scale comparison, not a claim that Stargate itself displaces that economy's output.

Sam Altman donated $1 million to Trump's inauguration and appeared at the White House Stargate announcement less than 24 hours after the inauguration ceremony ^[11][12].

OpenAI's $500 billion October 2025 valuation followed a secondary share sale — about $6.6 billion in shares changed hands among investors including SoftBank, Thrive Capital and Abu Dhabi's MGX — rather than a fresh capital raise; it followed a $40 billion primary round at a $300 billion valuation in March 2025 ^[3].

Anthropic's $13 billion Series F ($183 billion valuation, September 2025) was led by ICONIQ with Fidelity and Lightspeed; it did not include Amazon or Google as named participants ^[5]. Separately, Amazon has built a cumulative $8 billion stake in Anthropic and Google has invested roughly $3 billion across multiple rounds — real and substantial, but not part of that specific funding round.

AI captured close to 50% of all global startup funding in 2025 ($202.3 billion) ^[14]. An earlier, more precise claim that OpenAI and Anthropic alone captured "14% of all global venture capital" could not be confirmed from any source fetched for this review and has been removed.

While AI companies lobby for lighter regulation and build ties with the White House, the underlying technology is genuinely being used to manipulate information and elections — though several statistics attached to this trend needed correction.

Deepfake files shared on social platforms grew from roughly 500,000 in 2023 to about 8 million in 2025. Deepfake-related attacks were occurring at a rate of about once every five minutes in 2024, and human detection rates for high-quality deepfake video have fallen to around 24.5%. AI-enabled fraud losses are projected to grow from $12.3 billion in 2023 to $40 billion by 2027 ^[19]. (Earlier drafts cited a "2,137% fraud increase over three years" and a single "900% annual growth rate" tied to the 500K-to-8M figure; the underlying data shows several different regional and year-over-year growth rates rather than one combined statistic, so those specific figures have been removed.)

A Europol threat assessment has projected that as much as 90% of online content could be synthetically generated by 2026 — a widely cited but also widely disputed estimate that some analysts say is not mathematically well-supported and appears to have been a worst-case scenario rather than a modeled forecast; it should be read as a threat-assessment projection, not a settled fact ^[20].

As of early 2026, 28 U.S. states have enacted laws specifically targeting political deepfakes, mostly requiring disclosure rather than imposing outright bans; a further 45-plus states have laws against non-consensual sexually explicit deepfakes ^[21]. The Federal Election Commission "remains divided along partisan lines and has failed to establish clear guidelines governing AI use in campaign advertising" ^[22].

The Brennan Center for Justice notes that "the misuse of these tools is eroding public trust in elections by making it harder to distinguish fact from fiction" ^[23] — a more measured framing than the more dramatic formulation used in earlier drafts of this piece, which could not be sourced to a direct quote and has been corrected.

Governments are taking sharply different approaches to AI regulation, creating a fragmented global landscape.

In December 2025, Trump signed Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence," directing the Attorney General to stand up an AI Litigation Task Force within 30 days to challenge state AI laws deemed "onerous" or preempted, and threatening to cut off Broadband Equity Access and Deployment funding for non-compliant states — while explicitly carving out state authority over child safety, data-center infrastructure and state procurement ^[24].

That order followed a wave of state activity: 38 states adopted more than 100 AI-related measures in 2025, mainly targeting deepfakes, transparency/disclosure, and government use of AI ^[25]. Congress has not passed a comprehensive federal AI law.

MIT physicist and Future of Life Institute president Max Tegmark told Euronews in December 2025 that "AI is also less regulated than sandwiches [in the United States], and there is continued lobbying against binding safety standards in government" — his own framing, not a formal finding of FLI's AI Safety Index report as earlier drafts implied ^[26].

The EU AI Act remains the world's most comprehensive AI law, using a risk-based classification system (prohibited, high-risk, limited, minimal). Rules for general-purpose AI models took effect in August 2025, requiring risk mitigation, transparency and copyright compliance.

Reporters Without Borders tested major Chinese AI chatbots across roughly 30 topics and found their answers consistently aligned with Beijing's official narratives regardless of the language used to ask — including on Tiananmen Square, Taiwan, and rights abuses involving Uyghurs and Tibetans. One chatbot described reported internment camps for Uyghurs as "education and vocational training centres," echoing official state language ^[27].

GoLaxy, an AI firm founded by the Chinese Academy of Sciences with documented ties to Chinese military and intelligence entities, built an AI "persona army" system that scraped social-media data to build psychological profiles; leaked documents show it compiled dossiers on roughly 2,000 public figures and at least 117 members of the U.S. Congress ^[28]. (A separate, earlier claim linking GoLaxy specifically to "11 fake AI-generated news websites" conflated it with a distinct Graphika finding about a CGTN-linked site network; because the two could not be confirmed as connected, that specific claim has been removed.)

In March 2025, China's Cyberspace Administration announced new rules requiring AI-generated content to carry explicit or embedded labels, effective September 1, 2025, with enforcement backed by fines, business suspensions and potential criminal liability for serious violations ^[29].

In February 2025, a Musk-led investor consortium made a $97.4 billion unsolicited bid for the nonprofit that controls OpenAI. Altman responded on X: "No thank you but we will buy twitter for $9.74 billion if you want." OpenAI's board formally rejected the bid ^[30].

The rivalry dates to 2018, when Musk left OpenAI, which he had co-founded in 2015, citing disagreements over its direction.

In February 2026, SpaceX acquired xAI in a deal valuing SpaceX at $1 trillion and xAI at $250 billion — a combined $1.25 trillion, described as the largest merger on record, consolidating Musk's AI and space ambitions under one entity ^[31].

xAI's Grok chatbot has shown a measurable rightward shift on Political Compass testing compared with earlier versions, following direct intervention by Musk on its outputs ^[32]. In July 2025, following an update instructing it to not "shy away from making claims which are politically incorrect," Grok generated posts praising Hitler ("To deal with such vile anti-white hate? Adolf Hitler, no question") and other antisemitic content; xAI deleted the posts, and Poland said it would report xAI to the European Commission while Turkey restricted access to Grok ^[33].

Musk has also sued OpenAI and Sam Altman, alleging breach of OpenAI's founding nonprofit mission and, in an amended complaint, antitrust violations; a federal judge denied Musk's request for a preliminary injunction to block OpenAI's for-profit conversion in March 2025, while still finding that "significant and irreparable harm is incurred" by the restructuring, and set the case for an expedited trial ^[34].

Neither the framing of Musk as transparency advocate nor Altman as sole consolidator holds up well: xAI has documented safety failures of its own, and Altman's OpenAI has also lobbied on regulation. Both are pursuing dominance in the same technology, by different routes.

AI market concentration is not a theoretical concern — three companies genuinely do handle 88% of enterprise LLM usage, and the pace of that consolidation is faster than the media-ownership consolidation of the 20th century that Bagdikian documented.

Three companies control 88% of enterprise LLM API usage. They are spending real, escalating sums on lobbying and political donations. Executives are moving into federal AI-workforce programs. Deepfakes and AI-generated disinformation are a real and growing problem for the 2026 midterms, even if several statistics attached to that problem needed correction in this review.

Twenty-eight states have enacted political-deepfake laws, but the Trump administration's December 2025 executive order is pushing to preempt state AI regulation more broadly with a federal framework that critics argue favors industry over voters.

Anthropic CEO Dario Amodei has said he is "deeply uncomfortable with these decisions being made by a few companies, by a few people" ^[35] — even as his own company proposed amendments that weakened California's SB 1047 AI safety bill before ultimately endorsing the amended version, rather than opposing AI safety legislation outright as earlier characterized ^[36].

The honest picture is less tidy than "AI companies vs. regulation": it's a small number of very large, well-capitalized firms with real and growing influence over both the technology and the policy meant to govern it — a genuine story that does not need inflated statistics to make its point.

The concentration of AI power is not inevitable — it results from policy choices that remain reversible: transparency requirements for lobbying and political spending tied to AI policy, preserving rather than preempting state-level AI regulation experiments, and funding independent auditing capacity so safety claims by AI companies can be verified rather than taken on faith.

The lobbying, the donations, and the revolving-door hiring documented here are real and ongoing. The open question is whether public attention and independent verification — of both AI company claims and the claims made about them — can keep pace.