Anthropic CEO Dario Amodei launched a blistering critique of OpenAI, accusing the company of “gaslighting” the public and engaging in “safety theater” regarding its newly signed contract with the U.S. Department of Defense (DoD).
In an internal memo first reported by The Information, Amodei described OpenAI’s public messaging as “straight up lies,” intensifying the deepest rift yet between the two AI giants.
The Core of the Dispute: “Any Lawful Use”
The conflict centers on the difference between “technical safeguards” and “contractual red lines.”
- Anthropic’s Rejection: Amodei confirmed that Anthropic walked away from an expanded $200 million deal because the Pentagon refused to put two specific prohibitions in writing: no mass domestic surveillance and no fully autonomous lethal weapons.
- OpenAI’s “Swoop”: Hours after Anthropic was designated a “supply chain risk” for its refusal, OpenAI signed a deal. Amodei alleges that OpenAI accepted the vague “any lawful use” framework that Anthropic rejected, but is now publicly claiming to have the same “red lines” to manage its reputation.
- “Safety Theater”: Amodei argued that OpenAI’s reliance on “layered protections” and “field engineers” is performative. He claimed the real reason OpenAI took the deal was to “placate employees” and maintain government favor, whereas Anthropic chose to “hold its red lines with integrity.”
Political Allegations: The “Donation” Factor
In a move that surprised many industry observers, Amodei’s memo also touched on the political climate under the current U.S. administration.
- Campaign Contributions: Amodei suggested that the government’s hostile stance toward Anthropic—including President Trump’s “left-wing nutjobs” comment—was partially due to a lack of political alignment.
- The OpenAI Contrast: He pointed out that OpenAI President Greg Brockman and his wife had donated $25 million to a pro-Trump super PAC, implying that OpenAI’s deal was “greased” by political loyalty that Anthropic refused to provide.
Sam Altman’s Admission of “Sloppy” Optics
Facing an internal and external firestorm, OpenAI CEO Sam Altman has spent the week in damage-control mode.
- “Opportunistic and Sloppy”: Altman admitted on X (formerly Twitter) that the timing of the announcement—coming immediately after Anthropic was blacklisted—was “rushed” and “looked bad.”
- Amended Deal: On March 3, Altman announced OpenAI was amending its DoD agreement to add clearer, explicit language barring domestic surveillance of U.S. persons and usage by agencies like the NSA without further modifications.
- Philosophical Difference: Altman argued that OpenAI prefers a “layered approach” (technical stacks and embedded researchers) over the “legalistic” approach favored by Anthropic.
Market Consequences: The “QuitGPT” Shift
The public “war of words” has had immediate impacts on user behavior and app store rankings as of March 5, 2026.
| Company | Current Sentiment & Market Status |
| Anthropic (Claude) | #1 Free App on the U.S. App Store; Daily signups have tripled since the controversy began. |
| OpenAI (ChatGPT) | Facing a 300% surge in uninstalls; roughly 1.5 million Plus subscribers have reportedly canceled. |
| Google/Nvidia | Both companies signed an ITI letter urging the Pentagon to stop using “supply chain risk” labels as a weapon in procurement disputes. |
The Legal Front
Anthropic has officially confirmed it will sue the Pentagon to challenge its “supply chain risk” designation, calling the move “retaliatory and punitive.” Legal scholars suggest Anthropic may have a strong case, as the label is historically reserved for foreign adversaries like Huawei, not domestic firms with a contract disagreement.
