OpenAI’s Altman says Pentagon set ‘scary precedent’ binning Anthropic OpenAI has signed a deal with the United States Department of War (DoW) that allows use of its advanced AI systems in classified environments, and urged the Pentagon to make the same terms available to its rivals. The AI upstart revealed the deal in a Saturday post that said it includes the following three “red lines”: No …
No use of OpenAI technology for mass domestic surveillance.
No use of OpenAI technology to direct autonomous weapons systems.
No use of OpenAI technology for high-stakes automated decisions (e.g. systems such as “social credit”).
OpenAI fills the vacancy one day after Anthropic was evicted and maintains the same red lines Anthropic was evicted for?
Insert classic "I don't believe you" meme here ...
It was done as a 'pinky propmise' so legally enforcable as long as Altman and Trump remain BFF and are happy to share the same juice box.
If they fall out then all bets are off and the robots will be sent into the street to remove anyone saying the words "Epstein", "Orange toddler, "Cheeto in Chief" or "Taco".
"Why would the Pentagon seek to destroy one company over the same terms it agreed to with its largest competitor just hours later?
There’s an answer though: the words in OpenAI’s contract likely don’t mean what most people think they mean."
Well it does make you wonder why OpenAI says their agreement with the DoD is under supposedly the same conditions that Anthropic wanted and the DoD rejected.
I think we'll have to wait for the litigation to find out any details about exactly what Anthropic supposedly stopped DoD (and other agencies) from doing. Apparently one of the complaints was that Anthropic's rules weren't very clearly defined. And OpenAI's might not be much better, eg-
No use of OpenAI technology for mass domestic surveillance.
No use of OpenAI technology to direct autonomous weapons systems.
So define 'mass domestic surveillance'. DHS complained Anthropic imposed a similar restriction. There's also a complaint that DHS used IRS data to try and identify illegal immigrants, so whether that would be considered 'mass surveillance'. Which is also potentiall FUN! politically, ie regimes like California and NYC are facing a shrinking tax base, so identify high earners and offer them a pathway to citizenship. Or the US might want to identify fraudsters, but 'mass surveillance' restrictions might prevent government using AI to detect and prevent that.
And then there's hypocrisy and potentially useful AI guardrails. It would be amusing if government proposes the 'No AI for mass domestic surveillance' legislation.. Which might hopefully mean the end of Copilot, MS's automatic screenshotting, Google, Facebook and even OpenAI ingesting ever bit of PII it can get their grubby mitts on to profile every user.
The AWS restriction is also possibly open to interpretation. I doubt Ukraine & Russia are using OpenAI in their drone swarms, but AI is being used in that application. Which also depends on the definition of AI, eg US destroyers and GBAD is probably using some form of AI to detect and engage all the stuff flying around Iran at the moment. Probably still has a man-in-the-loop to set FCS to 'Auto', but in an era of drone, missile swarms and hypersonic weapons, defensive systems need fast decision making. If OpenAI or Anthropic want to shut themselves out of that lucrative market, it's their choice I guess.
Zyklon B ring any bells?
Why should it, or what's the relevance? Then again, drug dealers have been busily using 'AI' for rational drug design, and hopefully there are safeguards to stop those discovering better chemical agents.
IBM’s Hollerith machines might be more relevant, but profit often sadly trumps ethics.
Spot on Neil Barnes, my thoughts exactly:
"The AI System will not be used to independently direct autonomous weapons in any case where law, regulation, or Department policy requires human control"
Having worked in Policy, this is the biggest get out clause, where "law, regulation, or.... policy requires human control"... how long before the Tangerine Buffoon, or one of his hapless and feckless sidekicks, decides to change, law, regulation or policy.
In April, Fox News will "find" a memo that shows it was an AI that launched the strike on Iran, the one that started the by-now six weeks of ground war the US is involved in.
So, no, not tomorrow - two days ago.
Don't you recall DJT's post about the rogue AI : "He who controls the time stamps on the Truth Social server..."
I already see OpenAI endgame here. Financial colapse for the company is getting closer and closer, BUT if you are too big to fall or too important (defence contractor, without us China will win!), there will be nice and warm goverment bailout. Also Anthropic is getting nice PR stunt that "we are not a bad guy, see? we have morals" from this incident.
"The AI System will not be used to independently direct autonomous weapons in any case where law, regulation, or Department policy requires human control, nor will it be used to assume other high-stakes decisions that require approval by a human decisionmaker under the same authorities."
Or in other words all they need to do is change the law, regulation or policy to not require human control and they can use OpenAI's LLM to fire all the weapons and make all the decisions they want.
These promises mean nothing. Alman just eagerly signed up to build Skynet.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2026
