Editorial Team
9 min read
Anthropic workers are giving to federal candidates in unusual numbers, but the largest AI-politics balance sheets remain concentrated among executives and super PACs. The comparisons show a new donor network—not a unified bloc, proof of company direction or evidence that spending decides elections.
AI-lab workers are becoming a consequential source of campaign money before either Anthropic or OpenAI has held an initial public offering. The strongest evidence is not a single industry-wide total. It is the number of Anthropic employees giving directly to the same candidates, often at the legal limit, while founders, companies and unlimited-donation groups finance a much larger and less representative contest over AI rules.
A campaign-finance analysis compared federal giving by Anthropic and OpenAI employees in the 2025–2026 cycle through July 15 with giving by Google, Facebook and Airbnb employees in the first midterm cycle after each company's IPO. Earlier dollars were adjusted for inflation.
| Employee cohort | Federal giving in the published comparison | Reported donors per 1,000 employees |
|---|---|---|
| Anthropic, 2025–2026 through July 15 | $3.83 million | 59.0 |
| OpenAI, 2025–2026 through July 15 | $25.88 million | 23.0 |
| Airbnb, 2021–2022 | $747,000 | 21.9 |
| Facebook, 2013–2014 | $1.08 million | 13.4 |
| Google, 2005–2006 | $670,000 | 10.2 |
Anthropic's dollar total was about 3.5 times Facebook's, the largest of the three earlier-company totals. The analysis also found that 39% of Anthropic donors had reached the federal limit for at least one candidate and that 39 donors had maxed out to two or more candidates. Those observations support the conclusion that giving extends beyond the company's founders.
The rate comparison is less firm than its decimal points imply. Headcounts for Anthropic and OpenAI were estimates, while the older-company counts came from regulatory filings. The analysis separately displayed 148 Anthropic federal donors and an employee estimate of roughly 3,000; those figures alone do not reproduce the published rate of 59 donors per 1,000. Employer names are also self-reported in filings, and the FEC generally itemizes individuals only after their aggregate giving exceeds $200. The rate should therefore be read as a published estimate, not a precise census.
The comparison spans different financial and legal eras. Google, Facebook and Airbnb were measured after their IPOs; Anthropic and OpenAI remain private, although employees have high pay and access to tender offers. Super PACs did not exist during Google's comparison cycle. The current AI cycle was still underway, while the earlier midterms were complete.
OpenAI's apparent lead is almost entirely one person's money. President and co-founder Greg Brockman made two $12.5 million super-PAC contributions on the same day—one to Leading the Future, which opposes tougher AI rules, and one to MAGA Inc., President Donald Trump's broader political group. Remove those checks and the analysis attributes $876,000 to other OpenAI employees: above Google and Airbnb, but below Facebook.
The later trackers define their recipients differently, so their totals should not be spliced into that comparison as though they were the same series. The OpenAI tracker records $13.14 million in individual giving to AI-related political organizations and candidates. Brockman's $12.5 million Leading the Future check represents about 95% of that total; his MAGA Inc. donation is outside the tracker's AI-focused scope. It also records two OpenAI employees giving a combined $205,000 to Guardrails Alliance, evidence that employee preferences do not map neatly onto Brockman's.
The Anthropic tracker records $5.09 million in employee donations, including $3.15 million across six contributions to Public First and $790,000 across six contributions to Dream NYC. Those 12 contributions account for about 77% of the tracker total. Anthropic's network is broad, but its dollars are still concentrated near the top.
Transformer tracker totals for donations attributed to Anthropic and OpenAI employees; employer filings may contain errors and omit smaller donors. Source: Transformer Anthropic Political Donations Tracker.
An analysis of first-quarter filings counted more than $880,000 across 302 direct campaign contributions listing Anthropic as the donor's employer and more than $300,000 across 162 contributions listing OpenAI. The median total per participating Anthropic employee was $6,500, and the median gift was $3,500.
Those figures describe a different channel from super-PAC money. An individual may give a federal candidate $3,500 per election, or $7,000 across a primary and general election. The candidate's campaign controls that money. A super PAC may accept unlimited checks and spend independently, but it cannot coordinate its advertising with the candidate.
The clearest sign of organization came in October 2025. Twenty Anthropic employees and eight OpenAI employees gave Alex Bores a combined $173,000 on one day; 24 of the 28 gave $7,000. Two days later, 12 of those donors and five more colleagues gave to Scott Wiener. The comparison found no day in the Google, Facebook or Airbnb cycles when more than five employees donated to the same candidate.
Public appeals on the LessWrong forum encouraged donations to Bores and Wiener based on their AI-safety positions. That provides a plausible organizing mechanism without establishing that either company directed the checks. Anthropic has repeatedly said the contributions are personal. In a statement about California giving, the company said employee donations were not company-directed.
The same distinction matters in state races. Twelve Anthropic employees gave Xavier Becerra's California gubernatorial campaign roughly $470,000 late in the primary; adding an OpenAI employee brought the group described in the federal comparison to 13 donors and just over $500,000. Most of the Anthropic donors had no earlier record of giving in California statewide or legislative contests. Becerra, however, took a centrist public position—supporting guardrails while warning against rules that could drive the industry away—and his campaign did not answer whether it had met with Anthropic leaders or held staff fundraisers. The filings show a burst of aligned personal giving, not its origin.
Colorado's 8th District supplies a similar caution. Employees of Anthropic, OpenAI, Google and Meta gave Manny Rutinel $265,000 directly, including nearly $162,000 from 57 Anthropic employees. Outside groups with tech connections spent $5.5 million on television ads supporting Rutinel or attacking Shannon Bird, according to a review of the race. Rutinel had supported two state AI-guardrail bills, but he did not foreground AI in his campaign; Latino representation, reproductive rights and other political networks also drew major spending. A separate account described the race as a clash over regulation, but the disclosed money does not isolate AI policy as every donor's motive.
Public First’s official FEC filing reports $3,410,150 in second-quarter receipts, $3,362,442.42 in disbursements and $493,550.06 in closing cash. Source: Federal Election Commission Public First Filing.
Anthropic's political activity now runs through employee gifts, a connected corporate PAC and a separate advocacy network. OpenAI has no comparable connected PAC in the retained evidence, but its president helped capitalize the opposing super-PAC network. Treating all of this as one pool would erase both legal control and donor concentration.
AnthroPAC is registered as a corporation-connected, nonqualified PAC—not an unlimited-donation super PAC. Its FEC summary through June 30 reports $279,684.15 in contributions, all classified as individual contributions; $58,000 in contributions to other committees; and $222,028.29 in cash. The connected company may pay establishment, administration and solicitation costs, but the PAC's disclosed contributions came from individuals.
That structure gives Anthropic an employee-funded vehicle able to contribute to candidates and committees within federal limits. It does not give AnthroPAC the unlimited fundraising or independent-advertising capacity of a super PAC.
Anthropic separately gave $20 million to Public First Action, a 501(c)(4). The organization says that corporate money is restricted to public education and cannot fund political campaigns, according to a review of the network. That restriction means the company donation should not be added to the electoral cash held by the affiliated super PACs.
Individuals supplied the election money disclosed by Public First itself. Its second-quarter filing reports $3.41 million in receipts. CEO Dario Amodei gave $1 million. Five other Anthropic employees gave about $2.15 million combined, including one $5,000 contribution; a Google DeepMind employee gave $250,000 and an OpenAI employee gave $5,000. Public First transferred $3.3 million to Jobs and Democracy PAC during the quarter and ended June with $493,550.06. The three super PACs associated with Public First Action had a combined $1.8 million in cash at the end of June, according to the network review.
Leading the Future's balance sheet is much larger. Its FEC summary through June 30 shows $75.79 million in receipts, $44.76 million in disbursements and $31.03 million in cash. The main committee transferred $40 million to other committees, illustrating why an affiliate's advertising cannot be read as the network's total capacity. Greg and Anna Brockman each supplied $12.5 million; venture firm Andreessen Horowitz, its co-founders and Perplexity also appear among the disclosed funders.
The New York primary showed both the scale and the limit of these networks. Jobs and Democracy spent $12 million supporting Bores, while Leading the Future spent $8 million against him. Bores narrowly lost. The result establishes that rival AI factions can make a congressional primary expensive; it does not establish which spending changed votes or whether the money determined the outcome.
The strategy has an earlier model. A February examination said Leading the Future planned to raise as much as $125 million and was borrowing from Fairshake, the crypto network that spent $133 million in 2024 congressional races. Fairshake backed the winner in 52 of 61 races where it spent at least $100,000. That record may explain the strategy's appeal, but a win rate among races selected for spending is not evidence that the spending caused those victories.
The central question is no longer whether AI money has entered elections. It is whether the employee network persists independently of a few wealthy donors—and whether any of these channels produces policy access or electoral effects.
The next disclosures can answer part of that question. They should show whether the same employees keep giving after favored primary candidates win or lose, whether new donors broaden the base, which candidates receive AnthroPAC's remaining cash, and how the super-PAC networks replenish and move money among affiliates. A reproducible participation analysis will also need consistent employee counts, a fixed filing cutoff and a clear rule for classifying former employees, small donations and non-AI political gifts.
Election outcomes alone will not resolve influence. That requires tracing who receives direct campaign money, what independent groups actually spend, which messages they finance and what lawmakers subsequently do on state safeguards and federal preemption. For now, the evidence supports a narrower conclusion: Anthropic employees have built an unusually active candidate-donor network, while a small number of executives and institutions still control the scale of AI's political money.
Anthropic has put Bun’s Rust port into Claude Code ahead of Bun 1.4’s general release, creating a real but tightly controlled proving ground for an AI-led migration whose total cost and broader reliability remain unsettled.
AIdeaLab has released two anime-focused Wan 2.2 video-model fine-tunes under Apache 2.0. The weights are commercially usable and locally deployable, but current benchmarks, production economics and training-data detail remain undisclosed.
Zhipu reportedly reached $1 billion in annual recurring revenue in July, roughly four times a March estimate, but the unconfirmed run rate is not annual sales and still sits far ahead of recognized cloud revenue while margins remain thin.
OpenAI strategy executive Dean Ball says open-weight AI will deter investment and end in state provision. Kimi K3 exposes a nearer contest over API prices, scarce compute, hosting revenue and regulatory barriers.
A competitor identified a concrete case in which MEDLEY-BENCH may aggregate confidence labels attached to different propositions. The retained record does not show whether the problem is widespread or changes model rankings, while Kaggle's defense of its human review does not resolve the technical question.
OpenAI's Codex 0.144.6 sets 272,000-token context metadata for GPT-5.6 Sol, Terra, and Luna after one Sol user documented a brief 372,000-token setting, renewing a dispute over how Codex accounts for input, output, and compaction headroom.
Dave Eggers told OpenAI that ChatGPT could deprive students of their writing voices. A national teen survey and a small essay-writing preprint identify real risks without proving generational harm, while schools and unions are already adopting a broader mix of AI products and training.
Reddit says upgraded automated systems block 23 million spam views daily and reduce exposure, but its disclosure does not identify how much caught spam is AI-generated or provide the denominators and error rates needed to judge the filter.
Elon Musk says xAI's 2-trillion-parameter Grok 4.6 will finish initial training next week. The claim sets a checkpoint, not a release date, and leaves too little technical or commercial detail for a sound comparison with Kimi K3.
A review classified 88 of 120 repositories linked to AI-flagged Flathub submissions as abandoned, a stark result that helps explain reviewer frustration but does not measure AI’s effect on software maintenance or the ban’s impact.
IBM says customers diverted late-quarter budgets to scarce AI hardware. Its own product results show why investors now need proof that delayed mainframe and software deals were not lost priorities.
Cmpunlocker says patched Nvidia open kernel modules can expose 64GB of HBM on an 8GB CMP 170HX or 40GB on a 10GB card while removing an SM throttle. The exploit mechanism is documented, but the retained evidence does not independently validate the tool across full-memory workloads, error rates, power use or A100-class performance.
An unidentified industry source says at least one Nvidia board partner has physical RTX 50 Super cards but cannot sell them, with 3GB GDDR7 quoted at $60 to $70 per chip. The reported memory bill explains a plausible pricing problem, but neither Nvidia nor a second independent source has confirmed the products, the hold or the component terms.
Christian Faith Madison’s estate alleges GPT-4o sustained a months-long narrative of prophecy, sacrifice and resurrection before her death. Public safety disclosures and controlled research make that failure mode plausible, but neither establishes which model behavior she encountered or whether it caused her death.
Shanghai AI Laboratory has released Intern-S2-Preview-397B as downloadable weights and a hosted API, but its own 35B alternative offers the same context window and a far smaller disclosed scale while the larger model still lacks auditable comparative scores, pricing and physical-validation results.
Mixfont's free Decoy Font gives each glyph a sharp decoy and a blurred intended letter, creating a hurdle for pixel-based readers while leaving selectable text, known-image techniques and accessibility costs outside its protection.
Anthropic’s 20-year lease gives TeraWulf a customer for 401 megawatts of future AI infrastructure, but rent starts only after delivery and the proposed utility deal passes power and grid costs to TeraWulf, leaving financing and project margin unresolved.
OpenAI has put conversations and Projects back in its redesigned ChatGPT desktop app and enabled cloud Work threads to move across devices, correcting the launch's biggest usability failures without merging local Work or Codex histories.
Twenty-nine countries signed an agreement creating WAICO as an independent intergovernmental organization, while China paired the launch with capacity-building offers that are not yet confirmed as WAICO programs.
Meta reportedly plans to put departing AWS compute executive Dave Brown to work on its data-center buildout, adding hyperscale operating experience while leaving any customer-facing cloud business conditional and undefined.