Trump's AI Farce: Insult if You Don't Pay

Anthropic CEO Dario Amodei sent a 1600-word internal memo to all employees last Friday. The memo was leaked by the tech media outlet The Information today, setting off a bomb in Silicon Valley.

The core of the memo was just one sentence: Anthropic was blacklisted by the Trump administration not because they couldn't agree on security terms, but because they didn't donate.

25 Million Dollars Away

Dario specifically called out OpenAI in the memo.

He mentioned that the real reason the Trump administration didn't like Anthropic was that the company hadn't donated to Trump, hadn't offered "dictator-style praise." On the other hand, OpenAI had provided both funds and posture.

In September 2025, OpenAI President Greg Brockman and his wife donated $25 million to the Trump MAGA Inc Super PAC. According to Federal Election Commission records, this donation was the largest single donation to MAGA Inc for that period, accounting for nearly a quarter of the fundraising total for the half-year. Brockman later posted on social media, stating that the donation was to "support policies that promote American innovation" and praising the Trump administration for being "willing to engage directly with the AI community."

CEO Sam Altman took a different path. He didn't directly donate a large sum to MAGA Inc, but in December 2024, he donated $1 million to the Trump inaugural committee. More importantly was his posture: on Trump's second day in office, Altman stood behind the presidential seal at the White House's Roosevelt Room and announced the $500 billion AI infrastructure project Stargate, telling Trump in front of cameras, "This wouldn't be possible without you." At the White House tech dinner in September 2025, he faced Trump and said, "Thank you for being such a pro-business, pro-innovation president."

Interestingly, this same Sam Altman publicly wrote in 2016, "For anyone familiar with 1930s German history, watching Trump's actions is chilling." He even compared Trump to Hitler and discussed the "big lie." Before the 2024 election, he donated $200,000 to the Democratic Party to help Biden's reelection.

Meanwhile, Anthropic did nothing. No donations, no attendance at dinners, no standing behind the presidential seal to express gratitude.

This is the first time a CEO of a top AI company has publicly said this: your treatment in Washington depends on how much money you've given to the White House.

Dario also directly criticized the contract OpenAI signed with the Pentagon. He said the reason OpenAI accepted the contract was "they care about appeasing employees, and we care about genuinely preventing misuse," and called OpenAI's PR narrative around the contract "blatant lies."

The White House's response indirectly confirmed Dario's claims. A government official told the news site Axios, "You cannot believe Claude wouldn't secretly carry out Dario's personal agenda in a classified environment." Treasury Secretary Scott Bessent directly pushed back, stating that no private entity may dictate U.S. national security terms.

The response did not mention a word about the technical disagreement on security terms. It was all ad hominem attacks.

Following the leak of the memo, former Google CEO Eric Schmidt publicly sided with Dario, saying, "Dario is right; this is one of the most critical decisions our society faces." At the same time, OpenAI CEO Altman himself admitted in an internal memo that signing the Pentagon contract just hours after Anthropic was shut down "looks both opportunistic and hasty."

In a Double Bet Situation: Palantir's Awkward Position

The other half of the memo dismantles Palantir, America's largest defense data analytics company, with a market value of about $350 billion.

Palantir does not build models, does not make chips; it is like a government-specific OpenClaw, serving as an intermediary layer to plug in large models like Claude, GPT into your chat tools and workflows. Palantir integrates others' AI models into the military's classified data systems, allowing these models to read intelligence, run analysis, and perform target identification. 54% of its revenue comes from the government, with the U.S. market representing 74% of total revenue. Its relationship with the Pentagon is not a cooperative one; it is symbiotic.

By the end of 2024, Anthropic entered the Pentagon's classified network through Palantir, and Claude became the first cutting-edge AI model deployed in the U.S. military's classified systems. In July 2025, the Pentagon awarded Anthropic a $200 million contract. The Maven intelligent system operated by Palantir (the Pentagon's flagship AI project for intelligence fusion and target identification) contract's ceiling is mentioned at $1.3 billion. The U.S. military's six combatant commands, NATO, all run on Claude.

Then things started to go wrong. In January 2026, during a U.S. military operation to capture Venezuelan President Maduro, Claude assisted in intelligence analysis through the Palantir platform. Afterward, Anthropic inquired Palantir whether Claude was used in the firing process. Palantir then passed this question on to the Pentagon. The military saw this as the AI vendor retrospectively reviewing a military operation, irreparably damaging the relationship.

At a critical point in negotiations between Anthropic and the Pentagon, Palantir added fuel to the fire by pitching a proprietary "classifier" security solution to the Pentagon, claiming it could automatically determine through machine learning whether Claude's every use crossed a red line. The implication was clear: even if Anthropic refused to sign an unrestricted contract, we could control its model ourselves. This solution effectively provided the Pentagon with an out—since Palantir claimed it could manage it, Anthropic's security provisions were deemed unnecessary.

In a memorandum, Dario tore apart this proposal. He said, "About 20% is real, 80% is a façade." His reasons were as follows: the model could not determine if it was embedded in an autonomous weapons system loop; it was unclear if the data being analyzed originated from foreign sources or U.S. citizen data; it was uncertain if the data was obtained with user consent or through gray-market channels; jailbreak attacks were frequent and easily achievable. For these four questions, Palantir's classifier had no answers.

Dario said that Palantir's true understanding of Anthropic's position was: "You have some unhappy employees, and you need to give them something to appease them."

On March 3, Palantir CEO Alex Karp, speaking at the Washington Defense Technology Summit hosted by top Silicon Valley venture capital firm Andreessen Horowitz, took a veiled shot: "If Silicon Valley thinks they can take away everyone's white-collar jobs and then screw over the military, you are retarded." Everyone knew who he was referring to. What he didn't mention was that the "uncooperative" company in his rant happened to be his platform's most essential AI supplier.

Palantir sold an AI security layer to the Pentagon, powered by Anthropic's Claude AI. Anthropic's CEO said the security layer was a show. Palantir helped the Pentagon find a reason to kick out Anthropic, but after kicking out Anthropic, Palantir itself felt the pain.

Unplugging the Engine Is More Painful Than Imagined

Naturally, one might think: Claude is just a model base, can't we just switch to OpenAI's GPT or xAI's Grok? Like switching the default model in OpenClaw?

It's not that simple. Reuters today quoted two sources as saying: There are a lot of prompts and workflows in the Maven intelligence system built around Claude. This is not a matter of changing an API address. Prompt chains, workflows, output formats, security audit processes, all have been tuned to Claude's behavior patterns. Switching models means rebuilding and testing a complete process for military intelligence analysis and target identification. Sources said Palantir needs to "rebuild part of the software."

The Maven contract has a ceiling of $1.3 billion, extending until 2029. The deployment scope spans the six major combatant commands of the U.S. military and NATO. The timetable for the National Geospatial-Intelligence Agency, the primary U.S. geographical intelligence agency, is for Maven to start delivering "100% machine-generated" intelligence to theater commanders by June 2026. Now that the engine is changing, this schedule is highly likely to slip. An analyst from Wall Street investment bank Piper Sandler said: Anthropic is deeply embedded in the military and intelligence system, "access and negotiating alternative technologies takes time and resources, resources that could have been used on growth opportunities."

The prototype of the movie "The Big Short" and Palantir's most famous short seller, Michael Burry, added: The six-month transition period precisely illustrates that the stickiness is in Claude's technology, not on Palantir's platform. If Claude could easily switch models like in OpenClaw, why would the Pentagon need a six-month transition period?

Wall Street doesn't care about this. After Anthropic was blocked, tech boutique investment bank Rosenblatt raised Palantir's target price from $150 to $200, and UBS also upgraded its rating. On March 4th, Palantir's stock price rose by 3.28%. At the same time, CEO Karp and co-founder Peter Thiel sold over $400 million in Palantir stock between February 20 and March 3. Analysts are shouting "buy," while founders are selling.

On the same day the memo was leaked, there was another twist of events.

During Morgan Stanley's Tech Conference on March 4, Dario told investors that Anthropic was in talks with the Pentagon to "cool off and reach an agreement that works for both parties." He said Anthropic and the Pentagon have "far more in common than differences." According to sources, during the five-day blackout, Anthropic executives privately expressed regret to the Pentagon for their previous communication approach.

However, the leak of the memo may have complicated matters again. Axios reported that White House officials believe Dario's attacks on the Trump administration in the memo "could blow up the reconciliation opportunity." A government official's exact words were: "You can't trust Claude not to covertly push Dario's personal agenda in a classified setting."

Interestingly, OpenAI is also lending a hand. Altman, while signing the Pentagon deal, actively requested the government to "provide the same terms to Anthropic" and publicly opposed labeling Anthropic as a "supply chain risk." He called it a "very bad decision."

The Pentagon had 48 hours to make a decision on Anthropic. The wait for Palantir's engine teardown has now exceeded a week. And now, the two sides are back at the negotiating table.