A look at the rights AI companies have in US government contracts, such as the “any lawful use” standard, amid the Anthropic-DOD dispute and the OpenAI-DOD deal
But Users Aren't Buying It
Jessica Tillipman
Context & Ripple Effects
The dispute over Anthropic’s defense-contract terms followed the DOD’s public designation of the company as a supply-chain risk, a move whose consequences for Anthropic and its customers were already under review in the designation’s contract fallout.
OpenAI’s separate arrangement has put the practical meaning of contractual safeguards under scrutiny: reporting says it accepted compliance with US law while the DOD held firm on bulk-data analysis demands in its DOD negotiations. This legal framing matters because users are questioning whether stated limits meaningfully narrow permitted uses.
First-order effects
Anthropic and the DOD must resolve their dispute through the rights and obligations embedded in government-contract language, rather than solely through each side’s public AI-use principles.
OpenAI’s DOD deal becomes an immediate comparator: its commitments will be judged against the breadth of an “any lawful use” standard and the uses government law permits.
Second-order effects
Other AI vendors pursuing defense work will have to examine whether their safety policies, data-handling promises, and contract clauses can coexist, or whether they need to price and negotiate for narrower terms.
Enterprise customers and partners may treat the Anthropic designation and contract conflict as a continuity risk, increasing the value of clarity on government-use rights before adopting a provider.
Third-order effects
If broad lawful-use clauses remain the default, procurement terms could become the decisive governance layer for dual-use AI—potentially overriding more restrictive vendor policy language in practice.
The conflict may force a clearer market split between providers willing to accept expansive government-use terms and those seeking enforceable deployment limits; the extent of that split remains contingent on future contract outcomes.
The trend: This is part of the shift toward sovereign AI procurement, in which government contracting language increasingly determines how frontier models can be deployed.
For the avoidance of doubt, the OpenAI - @DeptofWar contract flows from the touchstone of “all lawful use” that DoW has rightfully insisted upon & xAI agreed to. But as Sam explained, it references certain existing legal authorities and includes certain mutually agreed upon safe…
A lot of the concerns about the government's “all lawful use” language seem to stem from mistrust that government will follow the laws. At the same time, people believe that Anthropic took an important stand by insisting on contract language around their redlines. We cannot
From reading this and Sam's tweet, it really seems like OpenAI *did* agree to the compromise that Anthropic rejected - “all lawful use” but with additional explanation of what the DOW means by all lawful use. The concerns Dario raised in his response would still apply here
the contract snippet from the openai dow blog post is so obviously just “all lawful use” followed by a bunch of stuff that is not really operative except as window dressing. the referenced DoD Directive 3000.09 basically says the DoD gets to decide when autonomous weapons systems
Lots of new, hard to follow details today about the OpenAI-Pentagon deal. Here's a roundup of the most important things about using commercially available data for surveillance on Americans. TL;DR: It seems the Pentagon wanted Anthropic to allow this, and Anthropic's refusal is
For those wondering how mass domestic surveillance could be consistent with “all lawful use” of AI models, I recommend a declassified report from the ODNI on just how much can be done with commercially available data (CAI): “...to identify ever person who attended a protest” [ima…
OpenAI has released the language in their contract with the DoW, and it's exactly as Anthropic was claiming: “legalese that would allow those safeguards to be disregarded at will”. Note: the first paragraph doesn't say “no autonomous weapons”! It says “AI can't control [image]
My thoughts on OpenAI's agreement with the DoD: On autonomous AI weapons: 1. “The AI System will not be used to independently direct autonomous weapons in any case where law, regulation, or Department policy requires human control.” This says that OpenAI's models may not [image]
“We cannot say that the government cannot be trusted to interpret laws and contracts the right way, but also agree that Anthropic's policy redlines, in a contract, would have been effective.” This is a fair and good point.
Confirmation by the administration that the OpenAI contract contained the “all lawful use” wording that Anthropic rejected. Sam's wordsmithing aside, this opens the door for Trump or a future leader to authorize autonomous weapons or mass domestic surveillance with AI.
There is a narrow but important gap between the “all lawful use” stipulation and “no autonomous weapons.” On the one hand, you could interpret these two positions as being essentially aligned. But it is more complicated than that. 🧵
I feel like I am going insane and no one has read the articles. It appears that OpenAI has not brought about harmony and still has the “all lawful use” clause in their contract that was the issue in the first place? I think they've negotiated functionally the same contact they've
What we know about the OpenAI-DoW deal: OpenAI agreed to the terms Anthropic rejected. The terms include an “all lawful use” clause. The contract “references certain existing legal authorities” which the govt claims prove that domestic mass surveillance is already illegal.
@tedlieu The axios article doesn't have much detail and this is DoW's decision, not mine. But if the contract defines the guardrails with reference to legal constraints (e.g. mass surveillance in contravention of specific authorities) rather than based on the purely subjective co…
In fewer words: Anthropic doesn't trust the current administration's own interpretation of “all lawful use” and wanted consultation. OpenAI was more than happy to trust Hegseth and Trump with their technology.
Anyone who thinks “all lawful use” + LLMs doesn't enable unprecedented mass surveillance is ignorant of the state of the law, the state of the technology, or both.
This is just “all lawful use” with extra words - no way the pentagon would have a huge hissy fit about these redlines and then immediately agree to a new contract with the same ones in it
There's some discussion about whether contract terms ("all lawful use" vs more specific terms) vs safety stack (monitoring systems) are more effective as safeguards against AI misuse. It'd be useful for someone to game out how they'd hold up against historical incidents of
Can AI companies restrict government use of their technology? They do it all the time. Whether and how depends on the acquisition pathway, contract type, and terms. My explainer: https://jessicatillipman.com/ ... #Anthropic #openai #pentagon #DoD #govcon
In case anyone's interested, @jtillipman.bsky.social has an excellent, detailed analysis of the current Anthropic-DoD-OpenAI contract debate - lots of nuances I wasn't aware of! — #USPol #AI #AIGovernance #Anthropic #DoD #OpenAI #GovernmentProcurement #GovCon #ProcurementPolicy…
A much more wholistic analysis of the OpenAI v Anthropic v DoW contract mess — OpenAI gives up contractual enforcement of redlines in exchange for architectural enforcement (supposedly) — the incident highlights severe problems with government procurement — jessicatillipman.c…
An actual expert on government contracts: “Contractors restrict the government's use of their products all the time.” — Ben Thompson: “this insistence on controlling the U.S. military, however, is fundamentally misaligned with reality” — I just don't know who to believe!