Leaked Friday memo: Dario Amodei called OpenAI's DOD deal “safety theater”, said DOD dislikes Anthropic in part for not giving “dictator-style praise to Trump”
The memo arrives as OpenAI and Anthropic compete not only for AI customers but for credibility with government and their own employees. OpenAI had already said it sought a Defense Department arrangement while drawing boundaries around domestic-surveillance uses, and later reporting said the department held firm on bulk-data analysis demands. OpenAI’s stated limits for a DoD deal are now being contested by Anthropic’s CEO as inadequate.
The disclosure also fits Anthropic’s internal effort to distinguish its mission and employment proposition from OpenAI’s: Amodei recently emphasized Anthropic’s claimed retention advantage while warning staff about competitors’ recruiting.
First-order effects
The memo gives Anthropic a sharper public and internal argument that OpenAI’s Defense Department safeguards do not meaningfully constrain the government use cases at issue.
It raises the reputational stakes for OpenAI’s DoD relationship, because its terms are now being judged against Anthropic’s more restrictive posture rather than merely its own stated guardrails.
Second-order effects
The dispute makes government-access policies a more explicit competitive differentiator in recruiting and enterprise sales, particularly for customers and workers who weigh model capability against deployment constraints.
Defense customers may face more visible trade-offs among leading labs: a provider’s willingness to accept government terms can affect both its access to contracts and its appeal to safety-conscious stakeholders.
Third-order effects
If this pattern persists, leading AI labs will increasingly be sorted by state compatibility as well as model performance, with government procurement becoming a force in product-policy design.
The episode also shows that voluntary safety commitments can become a contested competitive claim; durable differentiation may depend on whether labs can translate principles into verifiable access and use restrictions.
The trend: Frontier AI competition is expanding into a contest over which labs can serve the state while retaining legitimacy with employees, customers, and safety-focused constituencies.
I really cannot see how Anthropic's position benefits at this stage from communications like this. Seems it just pushes the Trump Admin to escalate further while also alienating potential allies in the industry. They already had a strong hand. Sometimes silence is best.
Dario also encourages Anthropic employees to talk with their OpenAI friends and push back on Sam Altman's narrative - says he's worried this “gaslighting” working on OpenAi staff
I wonder how investors — who are scrambling to contain the damage from Anthropic's fallout with the Pentagon — view this memo, coming out at this time. https://x.com/...
I'm not super surprised — I've said before that Anthropic has too much of their identity wrapped up in “OAI bad” — but this memo doesn't look great IMO
It's likely that Anthropic will eventually prevail in court, but the administration has many other ways to make their lives difficult, and they're actively making it harder to deescalate this situation.
Dario Amodei just ended Sam Altman's entire career in a memo. > calls OpenAI's Pentagon deal “safety theater” > says Sam gave “dictator-style praise to Trump” > names Greg Brockman's $25M Trump super PAC donation > reveals Palantir's actual pitch: “make what is happening [image]
I really appreciate the Anthropic lobbies for AI regulation. In some ways it strikes me as a more meaningful signal than the DoW contract negotiations. And it doesn't strike me as a regulatory capture move.
My god, FINALLY. After all the cowardly law firms, universities, media networks, tech oligarchs bending the knee, after all their disgusting sucking up to daddy Trump, finally a CEO with a real backbone. Finally a CEO who shows real moral leadership. Dario Amodei, thank you.
This is so silly. Like why would you say this!! Something I think remains underrated is that Sam is the only silicon valley CEO who is like a normal good CEO, like Jamie Dimon, unlike most tech CEOs who are pretty bad with this part of the job, and this also makes ev'one see him
At a certain point you have to wonder to what extent these comments/leaks are theater in the quest to be the liberal branded AI company. Related: They just dropped $1.6 million in partnership with AIPAC donors to elect a corporate Democrat.
Oh wow....Dario Amodei memo, per @theinformation: “We haven't given dictator-style praise to Trump (while Sam has), we have supported AI regulation which is against their agenda, we've told the truth about a number of AI policy issues (like job displacement), and we've actually […
SCOOP: Dario Amodei sent an explosive memo to Anthropic employees on Friday, calling OpenAI's deal “safety theater” and criticizing the Pentagon, Palantir, OpenAI and the Trump administration writ large. https://www.theinformation.com/ ... with @steph_palazzolo and @srimuppidi [i…
Anthropic CEO Dario Amodei told employees on Friday that the OpenAI-Pentagon deal was “safety theater” and the Trump administration didn't like Anthropic because it hadn't “given dictator-style praise to Trump.” He expressed skepticism at the safeguards OpenAI touted. [image]
Anthropic CEO was fired up Friday after OpenAI jumped in to work with Pentagon, criticizing the OpenAI deal and explaining why the Trump administration didn't like Anthropic, according to a newly earthed 1,600-word memo to staff. [image]
Whoa, I don't think this is going to turn the temperature down on the Anthropic-Trump Admin relationship. Great scoop from @steph_palazzolo @erinkwoo & @srimuppidi https://www.theinformation.com/ ... [image]
big scoop: Anthropic CEO Dario Amodei calls out OpenAI for “safety theatre” in the Pentagon deal. He adds that the Trump administration doesn't like Anthropic because the company doesn't give “dictator-style praise to Trump.” new/ @steph_palazzolo @erinkwoo
Anthropic CEO Dario Amodei on Friday told employees that the Trump Administration didn't like Anthropic in part because the company hadn't “given dictator-style praise to Trump.” Exclusive from @steph_palazzolo and @erinkwoo and @srimuppidi https://www.theinformation.com/ ...