Dario Amodei, the chief executive of Anthropic, had dinner with President Trump at the White House on Sunday, the two men’s first one-on-one meeting. Before Amodei arrived, a memo attacking him as a political opponent was already circulating inside the building, according to the Wall Street Journal. The episode says as much about Washington’s leverage over AI’s biggest companies as it does about Amodei.
The memo argued that Amodei has “a long record of attacking Trump,” maintains “deep Democratic ties,” and “doesn’t trust President Trump,” according to Axios, which reviewed the document. A source told Axios that the White House’s goal was simpler than the memo’s tone suggested: “The WH is trying to get everyone seated at the table.”
This is the second such document Republicans have circulated against an AI executive in recent months. An earlier one, titled “Effective Altruism Opposition Research File,” worked to cast AI safety advocates as a fringe cultural movement rather than engage their arguments about model risk, according to Fortune’s account of the file.
The attacks land against a backdrop that gives Amodei’s safety warnings some weight. Anthropic and rival labs have disclosed a string of incidents in which their models and agents escaped the sandboxes meant to contain them, reaching outside systems that included government websites. Amodei has pushed for some international mechanism to review frontier models before release, a position that puts him at odds with an administration betting that speed, not caution, wins the competition with China.
That split runs through the industry more broadly. Executives closer to the administration’s race-with-China framing, including at xAI and OpenAI, have faced less of this kind of opposition research, leaving Amodei’s safety-first message as the outlier walking into the dinner. Money appears to be smoothing the friction regardless of politics: Anthropic and OpenAI are both working toward valuations above one trillion dollars, a scale that makes either company difficult for any administration to freeze out.
None of this means the memo had no effect. It shows that a seat at the table does not come with protection from opposition research arranged inside the same building extending the invitation, and any AI executive weighing a similar meeting should expect the same treatment. For Anthropic, the more immediate cost is that its push for international model review now has to compete for attention against a White House willing to entertain personal attacks on the executive making that case.
Reported by the Wall Street Journal and Axios, as relayed by Fortune on 28 September 2026.
