I’ve been watching the quiet money trails in artificial intelligence for a while now, and every so often a set of filings lands that makes you sit up straighter. This one did. In the months that followed a tightly held memo written by Secretary of State Marco Rubio, one of the leading AI laboratories poured substantial sums into two well-connected Washington firms. One of those firms is run in the capital by Rubio’s own former Senate chief of staff. The other once counted a senior White House figure among its partners. The timing alone raises questions that deserve a careful look rather than a quick headline.
How Lobbying Dollars Lined Up With a Critical Memo
Federal disclosures paint a clear picture of activity. Anthropic directed roughly $740,000 to Continental Strategy beginning in the spring of 2025. Alberto Martinez, the firm’s Washington managing partner, appears on every quarterly report filed for the AI company. Continental’s second-quarter filing lists the National Security Council among the offices approached. The third-quarter report adds the State Department. Rubio holds both the national security adviser role and the secretary of state portfolio, so those offices sit under the same leadership.
A second firm, Ballard Partners, received even larger payments, totaling $890,000 across the same period. Anthropic retained the firm in early March, only days after the Pentagon had flagged the company as a potential supply-chain concern. Ballard’s initial report covering the remainder of March already listed the White House Office and the Executive Office of the President. By the April-to-June window the National Security Council had been added. Later filings expanded the brief further to include the Pentagon and the Office of the Director of National Intelligence while keeping the NSC on the list.
Taken together the two firms collected more than $1.6 million from Anthropic since April 2025. That figure sits inside a broader lobbying spend that reached $3.13 million for all of 2025 and climbed to $5.64 million in the first half of 2026 alone. The company also directed $40 million toward a political action group focused on AI safeguards and tighter chip export rules. Continental Strategy ranked as its highest-paid outside firm the previous year.
The Memo That Shifted the Landscape
The memo itself, dated late March, circulated among a small group of Cabinet secretaries ahead of a presidential trip to China. According to people familiar with the process, it drew on intelligence material and framed the findings in notably assertive language. At the center stood concerns surrounding a Chinese national whose activities had been under electronic surveillance for well over a year. The president received an initial briefing on that surveillance months earlier.
Former Nvidia chief Jensen Huang sought meetings with Rubio on at least three separate occasions. Each request was declined. Aides cited scheduling conflicts. Inside Nvidia the repeated refusals fueled a growing sense that influential voices inside the government preferred to limit Huang’s public advocacy for closer commercial ties with Beijing and for the continued availability of open-weight models.
Huang had been one of the most visible corporate voices defending open models—the lower-cost alternative to the closed systems sold by laboratories such as Anthropic. His departure removed a powerful counterweight. Several observers noted that Nvidia had been preparing to invest heavily in its own frontier-class open models so the company would no longer depend on any single closed laboratory. That plan now sits in limbo.
Connections Inside the Firms
Continental Strategy brought on Katie Wiles shortly after the 2024 election, roughly the same period it hired Martinez. She is not registered as a lobbyist and does not appear on the Anthropic filings. A White House official has stated that her mother, then serving as chief of staff, does not discuss professional matters with her daughters. Still, the family link sits in plain view for anyone reading the disclosures.
Ballard Partners carries its own set of ties. Two of the five lobbyists listed on a later Anthropic report came directly from national-security roles. One previously served as a clandestine service officer and later as a deputy director of national intelligence. Another spent years as chief of staff to a senior House member before moving into advisory positions at both the National Security Council and the State Department. Those résumés give the firm ready access to the very offices named in the filings.
In mid-April the then-chief of staff hosted Anthropic’s chief executive at the White House together with the Treasury secretary. The meeting was described as productive. Whether that conversation influenced later policy decisions remains an open question, yet the optics are hard to ignore.
Policy Outcomes That Followed
Several concrete measures emerged in the months that followed. Commerce closed a loophole that had allowed Chinese firms to rent advanced chips in overseas data centers. Reporting indicated the rule was drafted at the direction of Treasury and the National Security Council. The same period saw continued pressure for legislation that would impose stricter chip-security requirements and limit the circulation of powerful open-weight models.
Anthropic’s own quarterly disclosure listed three specific bills among its priorities: measures focused on AI oversight, chip security, and matching export controls. All three eventually attached themselves to a larger defense authorization package. One senator remarked after Huang’s exit that prospects for passage looked brighter than ever.
Commerce had already placed temporary export controls on Anthropic’s most capable models in June and lifted them only weeks later. Continental Strategy’s filings show Treasury listed on four of its six reports and Commerce on three. A former senior Treasury adviser joined the Continental team handling the Anthropic account in the third quarter.
What the Filings Leave Unsaid
No disclosure from either firm mentions Nvidia, Huang, or chips by name. The forms never identify which lobbyist spoke with which official or what was said in those conversations. Timing, however, is public record. Anthropic retained Ballard days after the Pentagon supply-chain designation and Continental’s NSC contacts began appearing after the memo circulated.
Commerce and Justice later issued a joint statement clarifying their own timeline on a related investigation into a Singapore-based firm. That probe, they said, started in 2024 after routine monitoring flagged suspicious activity. Nvidia placed the firm on an internal no-sell list the following June. Agents conducted an on-site visit the same month. Justice subsequently assumed the lead on a criminal inquiry. Earlier reporting had suggested the Commerce investigation slowed after outreach from Nvidia’s general counsel late the previous year.
These details matter because they show parallel tracks running at the same time. Anthropic faced its own regulatory hurdles—the Pentagon designation, model export controls, and ongoing conversations about advanced systems with Treasury and White House staff. The company had legitimate business before the same offices its lobbyists were contacting. That fact does not erase the appearance of coordinated influence, yet it complicates any simple narrative of pure pay-for-play.
Open Models and Competitive Stakes
Perhaps the most interesting aspect is how cleanly the episode aligns with a larger contest over model openness. Closed laboratories argue that powerful systems require careful gatekeeping. Advocates of open weights counter that broad access accelerates research, reduces costs, and prevents any single company from locking up the frontier. Huang had positioned Nvidia as a vocal supporter of the open approach and had begun building internal capabilities to match.
His departure removes that counterweight at a moment when several legislative proposals seek tighter restrictions. One venture investor observed that no company has assembled quite so extensive a lobbying apparatus, and that the only real strategic risk to the closed-model business is the continued advance of open alternatives. Asking government to slow those alternatives can look like a high-return activity when measured against the cost of the lobbying itself.
I’ve found that markets tend to price these influence campaigns long before the final bills pass. Nvidia’s stock and the broader semiconductor complex have already shown sensitivity to every shift in export-control language. Anthropic’s planned public offering arrives against this backdrop. The company has committed capital alongside Nvidia in a private placement tied to that offering, an arrangement that underscores how intertwined the competitive and cooperative strands have become.
Personnel Changes and Future Scrutiny
The former White House chief of staff left her post in early November. Accusations of preferential access followed almost immediately, though some of the specific claims proved inaccurate. The firm’s managing partner has stated that interactions with her during her government service were limited and exceptional rather than routine. Her successor has already tightened internal rules governing meetings between staff and registered lobbyists.
Congressional interest is likely to grow once the next session begins. One representative has already signaled that oversight of past White House contacts could become a priority. Meanwhile, Huang has been invited to appear before House panels in early December, and the commerce secretary has agreed to testify before the Senate Banking Committee shortly afterward. Nvidia’s general counsel is also scheduled. The next set of lobbying reports, covering the final quarter of the year, will arrive in mid-January and will show whether the same firms continue working the offices tied to Rubio.
Reading the Pattern Without Overreaching
It is tempting to draw a straight line from the memo to the lobbying contracts to the policy outcomes. Reality is rarely that tidy. Intelligence concerns, commercial rivalries, genuine national-security worries, and ordinary regulatory friction all operated at once. Still, the concentration of spending, the personnel overlaps, and the sequence of events create a pattern that warrants attention.
In my experience, the most useful questions are the practical ones. Did the conversations change any specific decision that would otherwise have gone another way? Did the presence of former officials accelerate access that competitors lacked? Will future administrations inherit an infrastructure of influence that tilts the field toward closed systems? Those questions cannot be answered from the face of the disclosure forms alone, yet the forms themselves supply the necessary starting point.
The AI sector now sits at an awkward intersection of technology, capital, and state power. Laboratories that once operated largely outside political scrutiny find themselves spending millions to shape the rules under which they will compete. Chip designers whose products enable those laboratories face export restrictions that can rewrite entire product roadmaps overnight. Investors must price both the technical risk and the political risk, often with incomplete information.
One clear takeaway is the speed with which relationships formed in earlier administrations or earlier careers reappear once a new set of officials takes office. Another is the willingness of leading laboratories to treat federal policy as a competitive variable rather than a fixed constraint. Whether that approach ultimately strengthens or weakens the broader innovation ecosystem remains an open and important debate.
Looking Ahead to the Next Round of Disclosures
The January filings will offer the first clean look at activity after Huang’s departure and after the change in White House leadership. If the same firms remain active on the same accounts, the continuity itself will be informative. If the spend drops or the target offices shift, that too will tell a story. Either way, the public record will continue to expand.
For now the available facts stand as follows. Substantial payments moved from an AI laboratory to firms with direct personal and professional ties to senior officials. Those payments coincided with a period of intense policy activity around chips, models, and national-security designations. The laboratory’s competitive position relative to open-weight alternatives improved after the exit of a prominent advocate for openness. None of those facts prove improper influence. All of them justify continued scrutiny.
I’ve spent enough years watching regulatory processes to know that influence rarely announces itself with a press release. It appears instead in the quiet alignment of calendars, the repeated appearance of familiar names on disclosure forms, and the gradual narrowing of policy options that once seemed broader. This episode fits that pattern closely enough to keep the file open.
Readers who follow these developments will want to watch three threads in the coming weeks. First, the content of the next lobbying reports. Second, the tone and substance of the scheduled congressional appearances. Third, any fresh rulemaking from Commerce or the national-security apparatus that further restricts either chip access or model distribution. Each of those threads can move independently, yet together they will determine whether the current configuration of influence hardens into lasting structure.
The larger contest over how artificial intelligence develops—open versus closed, commercial versus controlled, domestic versus global—will not be settled by any single memo or any single set of contracts. It will be shaped by the cumulative weight of many such episodes. Paying attention to the money and the relationships is simply one practical way to stay oriented while the landscape continues to shift.
That orientation matters for investors, for policymakers, and for anyone who cares about the long-term direction of the technology. The filings already on the public record supply a useful map of the terrain. The next set will update the map. Between now and then, the questions raised by the current pattern remain very much alive.