(Look at Altman’s expression in this photo, it’s as if he’s saying: “Yes, actually I want it all.”)
Today is a rare sunny day in Shanghai, with temperatures around 6°C, feeling slightly cold. I was sitting by the window this morning, soaking up the sun and browsing the news, when I suddenly saw a particularly amusing piece of tech gossip.
This gossip perfectly illustrates what “top-tier corporate language art” is. You might have read countless press releases from tech companies claiming they are dedicated to “making the world a better place,” but today’s story is about how a super-giant elegantly and seamlessly swallowed its original intentions.
Boss, Did You Just Let That Slip?
A few days ago, OpenAI CEO Sam Altman held an all-hands meeting internally. The atmosphere was already quite tense because the company recently took on a military contract, and many employees had strong objections. As a result, in an attempt to pacify everyone, Altman casually dropped a bombshell.
He told employees that not only had they just secured a partnership with the Pentagon, but they were also considering deploying the company’s AI across all of NATO’s “classified networks.”
Not long after he said this, The Wall Street Journal leaked the story.
You can imagine how devastated the OpenAI PR team was when they saw the news alert. They probably dragged themselves out of bed in the middle of the night and rushed out a statement to various media outlets: “Everyone misunderstood! Our boss had a slip of the tongue; he actually meant NATO’s ‘unclassified networks’!”
This excuse is absolutely incredible. It’s like someone shouting in the street, “I’m going to rob a bank!” and his lawyer running over sweating profusely, covering his mouth, and explaining to the police, “He actually meant he’s going to the bank to open an account.” Anyone with a basic understanding of corporate PR can see that this so-called “unclassified” defense completely fails to cover up this Silicon Valley giant’s ambition to swallow a huge slice of the military-industrial complex pie.
Altman himself later admitted at the meeting that this was a “complex but correct decision,” even though it would put the company’s brand image in an extremely difficult position in the short term. Translated into plain English: The optics are indeed bad, but they are paying way too much to refuse.
A Rival Held the Line, Only to Get Blacklisted
Actually, OpenAI was able to secure this massive Pentagon contract so smoothly thanks to its peers making it look good.
Do you know why the Pentagon suddenly became so eager to approach OpenAI? Because just a few days prior, talks between them and another top-tier AI company, Anthropic, completely collapsed.
Anthropic is an anomaly in Silicon Valley. They established a very rigid “AI Constitution” and absolutely refuse to let the military use their technology for fully autonomous weapons or domestic surveillance. Their reasoning is also straightforward: current AI systems hallucinate all the time and are simply not reliable enough; using them in lethal weapons is far too dangerous and violates basic human rights.
They were incredibly stubborn. Rumor has it that just last month, Anthropic’s CEO specifically met with US Secretary of Defense Pete Hegseth to try and find a compromise. But the military was short-tempered and replied directly: Whether or not it is used in weapons is dictated by US law, and it’s not up to a private startup to dictate terms.
By March 3, 2026, the US government officially turned its back, designating Anthropic as a “supply chain risk”—essentially blacklisting them entirely. This forced Anthropic to run to a federal court in California yesterday (March 9) to sue the Pentagon.
(This scene is so dramatic: the one following the rules is forced into court, while the one breaking them is counting money)
Just as Anthropic was kicked to the curb for holding its bottom line, Altman and his team swiftly filled the gap, taking over the contract without missing a beat.
I was chatting with a friend who works on large models in Silicon Valley about this, and he dropped a hilarious piece of trivia.
Actually, before January 2024, OpenAI’s charter had a hard-and-fast rule: prohibiting the use of AI for military and warfare purposes. But later, they quietly deleted that sentence.
The best part is that before they even deleted the ban, the Pentagon was already using their models. How? Through Microsoft. The Pentagon simply bought Microsoft’s Azure cloud services because Microsoft has an exclusive license with OpenAI. The military didn’t even need to negotiate with a then-pretentiously-noble OpenAI; they just used the Microsoft backdoor. In other words, OpenAI’s overnight modification of its bottom line might not have been a carefully considered strategic pivot at all, but purely the realization that “Well, they’re already using it anyway, so I might as well make the money.”
Silicon Valley Was Not Like This Eight Years Ago
Watching Silicon Valley fight tooth and nail for military contracts today reminds me of another piece of history. Current tech giants might think embracing the military is a given, but the atmosphere eight years ago was completely different.
In 2018, Google took on a Pentagon contract called Project Maven, which mainly used artificial intelligence to analyze surveillance footage captured by drones, saving the military the hassle of watching videos manually.
Guess what happened? The moment this was exposed internally, Google employees absolutely exploded.
(The iconic scene of employees protesting back in the day; they even drew drone icons directly next to “Don’t be evil” slogans)
Over 4,000 employees co-signed a letter of protest, and dozens resigned on the spot. The engineers responsible for building the security firewalls for this project even resorted to “passive-aggressive slow-downs” to stall its progress. Everyone’s demand was incredibly simple: We are a tech company, not an arms dealer, and we cannot turn our technology into killing machines.
Google’s management at the time still cared about their reputation and was forced to yield by the employees. Not only did they announce they wouldn’t renew the contract with the military after it expired, but they also specifically released an “AI Moral Charter” pledging not to be evil.
Fast forward back to 2026. Taking on military contracts again: a rival is blacklisted by the government for sticking to a few ethical principles, while OpenAI’s boss is painting rosy pictures of classified NATO networks in internal meetings. In just eight short years, the guise of “technological neutrality” worn by tech giants has long been torn to shreds. No one even bothers to pretend anymore; they have thoroughly become the most eager cogs in the arms race.
The Dark Humor Behind the Name
Looking at the whole incident, what makes me sigh the most is the name “OpenAI” itself.
When the company was first founded in 2015, they solemnly pledged that it was a non-profit organization. They named it “Open” to keep artificial intelligence technology transparent, safe, and out of the monopolistic grasp of a few individuals or powers, for the benefit of all humanity ╮(╯▽╰)╭.
Looking back at that name now, and then looking at the classified Pentagon contract in their hands, along with a PR team desperately explaining “we were just browsing the unclassified networks,” it truly is top-tier dark humor.
Well, I’ve talked so much I’m getting hungry, I’m going to grab some lunch. Have you encountered this kind of “top-tier language art” at work today? Leave a comment and let me know.
References:
- OpenAI looking at contract with NATO, source says – Reuters
- Anthropic sues to block Pentagon blacklisting over AI use restrictions – Reuters
- OpenAI Quietly Deletes Ban on Using ChatGPT for “Military and Warfare” – The Intercept
- Pentagon Tested OpenAI Models via Microsoft Despite Military Ban – TechBuzz
- Google employees resign in protest against Air Force’s Project Maven – FedScoop
—— Lyra Celest @ Turbulence τ.
