On October 8, 2026, Anthropic announced the launch of its "Presidential Engagement Program," which will engage directly with presidential candidates from both parties during the 2028 U.S. election cycle, offer AI policy education, and prepare the company's executives for political conversations. The company simultaneously posted a job opening for a political program lead with an annual salary of $295,000 to $345,000, requiring at least 12 years of national political campaign experience and recognized credibility with both parties.
Over the same period, Anthropic has donated a cumulative $40 million to Public First Action: an initial $20 million in February 2026 and another $20 million in July, directed at AI safety guardrails, independent evaluation of high-risk models, and export control issues. CEO Dario Amodei personally donated $1 million to a super PAC that supports AI regulation. Internally, AnthroPAC has raised $372,000 in employee political contributions since its founding in April through August.
Why Now, of All Times
The timing is the product of three lines of pressure converging at once.
The first is open conflict with the Trump administration. In March 2026, the U.S. Department of Defense placed Anthropic on a danger list citing "supply chain risk," after the company refused to allow the Pentagon to use Claude for fully autonomous weapons systems or mass surveillance. Anthropic filed two federal lawsuits, and a California federal judge at one point ruled that the designation constituted unconstitutional retaliation for protected speech, but the D.C. Circuit Court of Appeals upheld the Defense Department's decision 2–1 on September 25, 2026.
The second is the approaching IPO. According to Fortune, Anthropic's expected valuation at listing could reach as high as $2 trillion. Any change in the regulatory environment will directly affect the company's pricing logic.
The third is the 2028 cycle window. By the rhythm of American politics, the real window for influencing the next administration's AI policy comes before the candidates take shape. Anthropic's job posting states: "The decisions made over the next two years will shape the future enormously. We want every serious candidate and elected official, regardless of party, to understand this technology and the case for governing it well."
What the Program Actually Is
Judging from the job description, the "political program lead" must map candidates across both parties, provide candidates and their advisers with "equal terms" AI policy training, prepare Anthropic executives for meetings with candidates, and lay groundwork ahead of party conventions, the drafting of campaign platforms, and the transition period.
Bruce F. Freed, president of the nonpartisan watchdog Center for Political Accountability, commented: "It is customary for corporations to maintain contact with presidential campaigns through their government relations departments. But I have not seen a presidential election engagement committee this formalized. My reading is that this is an AI company trying to shape or block government regulation." California State Senator Steve Padilla said: "Education and influence are not the same thing, and the public has a right to know where the boundary between them lies."
A Side-by-Side Comparison With Peers
According to CNBC, lobbying spending by OpenAI and Anthropic rose 23% year over year in the second quarter of 2026. OpenAI's federal lobbying spending reached $2.8 million in the third quarter, and it opened its first Washington office near the White House. Alphabet's federal lobbying spending hit $5.3 million in the second quarter of 2026 alone. AI-related lobbying spending totaled more than $1.2 billion in the first three quarters of 2026, and one in four federal lobbyists now focuses primarily on AI, compared with just 11% in 2023.
By comparison, Anthropic's $40 million donation to Public First Action stands out for the size of a single contribution, while the "Presidential Engagement Program" is more unusual in its organizational form — it is the first time among major AI labs that an internal political team has been set up for the next presidential election. Katie Harbath, former director of public policy at Facebook, noted that the role itself is not rare, but bundling presidential candidate outreach with management of a political action committee in a single position is indeed uncommon.
How Much Weight Can the "Education" Label Bear
Anthropic characterizes the program as "education," and the donation contract with Public First Action explicitly stipulates that the funds may not be used to influence the election of any federal, state, or local candidate.
Yet structural problems remain: on AI regulation, Anthropic advocates slowing the pace of development and building safety guardrails, which stands in direct opposition to the Trump administration's logic that "regulation will make America lose to China." When "educational" content is designed by a company team and delivered by political operatives recruited at high salaries, the line between "neutral education" and "framing" is hard to draw in practice.
Providing "equal terms" policy training to candidates of both parties does not mean the information is neutral. "Equal access" under any particular framework produces directional effects. Once a candidate's policy understanding is shaped by one company's vocabulary, subsequent policy advocacy is already partly accomplished.
The relationship between the current Trump administration and Anthropic has broken down, and with the Defense Department's "supply chain risk" label in place, the policy channel is all but closed. 2028 is the next policy window. Anthropic's move is a direct investment in the cognitive framework of the next administration.
Independent Judgment
The core logic of Anthropic's move is understandable: a company labeled a "supply chain risk" by the Defense Department, with litigation still underway, choosing on the eve of an anticipated $2 trillion IPO to systematically invest in the AI understanding of the next presidential candidates is a rational act of self-preservation.
The question is not whether the motive is self-interested, but the power asymmetry created by the scale and the form. $40 million in public education funding, a political operative at the $350,000 level, and systematic involvement in "presidential candidates' understanding" together constitute a new mode of political participation: the advance training of policymakers.
When a company's business model is deeply tied to regulatory outcomes, and it simultaneously possesses enormous funds, technical authority, and a particular set of positions, "education" naturally carries a direction. This is not a moral accusation but a structural description.
On the question of AI governance, who is qualified to play the role of "educator"? If it is a party with a direct stake in the market, the public has the right to demand transparency — not only disclosure of donation figures, but also disclosure of the "educational content" itself. Anthropic says it wants every candidate to "understand this technology," but whose framework of understanding is it, exactly? That is the core question.
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接