This is the AI platform aggressively promoted by the Pentagon; it actually looks as serious as a prime-time news broadcast.
It’s only about 6 degrees Celsius in Shanghai today. Although it’s sunny, the cold wind makes you just want to huddle indoors and scroll through your phone. I happened to stumble upon some incredibly surreal tech gossip that I just have to share with you all. Rubs hands together
The Ones Who Held Their Ground Got Eliminated First
Did you know? In the US, if you’re a tech company boss and you refuse a military contract because you don’t want your AI used for mass surveillance or installed on fully autonomous weapons, guess what happens?
You might think people would praise your principles, or at least respect your business choice. Wrong. The Secretary of Defense will straight-up label you a “supply chain risk.”
This isn’t some joke I made up; this is a true story that just unfolded in Silicon Valley over the past few days.
The unlucky protagonist of this story is Anthropic, the company behind the Claude large language model. Those familiar with them know that their founder, Dario Amodei, left his former employer, OpenAI, to start his own venture with a group of colleagues because he felt OpenAI was becoming too commercialized and not safe enough. From the very beginning, their main selling point was “ethical AI.”
So when the Pentagon came knocking, hoping to use their technology in military operations, Anthropic drew a hard line without hesitation: We absolutely will not allow our AI to be used for domestic mass surveillance, much less let it open fire without human intervention.
As a result, current Secretary of Defense Pete Hegseth was furious. He declared that since you don’t play ball, you’re a “supply chain risk.” He ordered a total ban that very day, effectively kicking them out of the group chat.
Altman’s “Teflon” Philosophy
Here comes the most surreal part.
Right after Anthropic got booted out, OpenAI swiftly signed this massive contract. Their moves were so fast, it almost looked like they had been waiting in ambush for a long time.
As soon as this happened, OpenAI’s internal staff exploded. After all, many employees had previously signed an open letter titled “We Won’t Be Divided,” showing real-name support for Anthropic’s backbone. But then their own boss turned around and picked up the meat the other guy threw away.
To appease these employees who were on the verge of walking out, CEO Sam Altman held a special internal all-hands meeting. His remarks truly took entrepreneurial pragmatism to its absolute extreme.
According to the leaked meeting transcript, Altman told employees very bluntly: “Maybe you think a strike on Iran is a job well done, while invading Venezuela is pure mischief. But it’s not your place to chime in on these matters.”
His logic is deeply realistic: deploying troops and firing weapons are government decisions; we are just tool sellers. Furthermore, if we refuse this job out of some sense of high-mindedness, Elon Musk’s xAI or other companies with zero safety concerns will take it. Things would only get worse then.
Sigh, the guy doesn’t have it easy. Being a CEO means constantly walking a tightrope between financial backers and employees.
When Altman was presenting his slides on stage, he probably didn’t expect a slip of the tongue later on.
Of course, just giving grand lectures isn’t enough; you still have to give everyone a graceful way out. So OpenAI’s PR team repeatedly emphasized that this collaboration is strictly for use on the military’s “unclassified networks.”
The Pentagon even built an internal system called GenAI.mil specifically for this, distributing it to three million Department of Defense employees. What is it used for? Mostly to help write documents and process tedious administrative approvals.
You can think of it as us just selling the military a bunch of kitchen knives to peel potatoes in the logistics cafeteria, absolutely not guns for the battlefield.
But this narrative was quickly slapped in the face. A Pentagon tech official turned around and told the media at a White House event that the military’s ultimate goal is to deploy these frontier AIs across networks of “all classification levels.” The funniest part is that even Altman himself slipped up at a conference, saying they were preparing to collaborate with NATO’s “classified networks.” This scared the PR team into full damage control mode, desperately explaining to reporters that the boss misspoke, and that it’s unclassified, unclassified.
Where Did the Backbone from Eight Years Ago Go?
Speaking of this, I am reminded of something else. When I first entered the industry writing tech news, around 2018, Google took on a Pentagon project called Project Maven. It was primarily about using AI to help the military analyze drone video footage.
When that news leaked, all of Google was in an uproar. Thousands of employees signed a protest petition, people furiously posted mocking memes on internal forums, and over a dozen engineers simply handed in their resignations. Diane Greene, the executive in charge of the cloud business at the time, couldn’t pacify them no matter what she did.
CEO Sundar Pichai simply couldn’t withstand the massive internal public pressure. He not only canceled the lucrative contract but also seized the opportunity to publish their famous “AI Principles,” solemnly promising never to use their technology for weapons development.
The scene of Google employees protesting the military project back then—that was truly hardcore.
Both situations involve dealing with the military. Eight years ago in Silicon Valley, if employees unfurled a banner, executives had to compromise. Fast forward eight years to today: OpenAI employees are protesting, but their executive directly fires back with, “it’s not your place to chime in on this.” I’m not sure if everyone’s mindset has just become more apathetic, or if the gravity of reality has just become too heavy.
Actually, there’s a fascinating piece of gossip dug up by the media. The Pentagon’s swift decision to kick out Anthropic and enthusiastically embrace OpenAI might be driven by a very direct factor beyond so-called commercial differences.
Rumor has it that OpenAI’s president, Greg Brockman, and his wife recently donated a whopping $25 million to a PAC supporting the current administration. Meanwhile, the uncompromising Anthropic hasn’t donated a single dime.
The way the business world operates can sometimes be so brutally simple it makes you laugh. Those tech ideals preached so beautifully at product launches truly look a bit flimsy in the face of $25 million in political donations and massive military contracts.
Alright, Time for a Bowl of Hot Noodle Soup
Oh, one more thing. When the Pentagon forcefully pushed this AI platform internally, the Secretary of Defense sent a mass email to all three million employees, encouraging them to use AI more to boost work efficiency. This scene feels incredibly familiar—it’s exactly like that admin supervisor who constantly nags you in the company group chat to clock in using the new attendance software. Except this guy holds the command of the armed forces ¯\(ツ)/¯.
I’ve been talking for a while, and my typing fingers are freezing. I need to go make a bowl of hot noodle soup to warm up. Have you guys ever experienced a situation at work where “the boss forcibly takes on a project you absolutely don’t want to do”? Feel free to vent in the comments; I’ll read them while I eat.
References:
- Sam Altman warns Elon Musk’s xAI could tell the Pentagon ‘we’ll do…’
- Sam Altman admits OpenAI can’t control Pentagon’s use of AI
- Exclusive-Pentagon pushing AI companies to expand on classified networks
- Google employees protest Pentagon partnership to CEO Sundar Pichai
—— Lyra Celest @ Turbulence τ.
