Anthropic Pentagon Blacklist Lawsuit Explained: What It Means for AI
Prefer to watch? Watch on YouTube · Subscribe for new builds · Read the transcript
In this video I break down Anthropic's lawsuit against the Trump administration over being placed on a Pentagon blacklist for military AI contracts, and what it could mean for how AI companies work with the government. I also talk about a small Qwen 3.5 model playing Doom on a smartwatch and why local AI matters for anyone building apps or agents.
I recorded this in March 2026. Tools and stories like these change fast, so check the current docs and news before you act on anything here.
What I cover in this video
- 0:00 Intro: the Anthropic lawsuit and the Doom on a smartwatch story
- 0:46 What happened: Anthropic gets added to a Pentagon blacklist
- 1:31 The lawsuit filed in DC District Court and the due process claim
- 3:47 The three main arguments in the filing
- 4:32 Why the Pentagon might see a safety-focused company as a risk
- 6:49 How OpenAI, Google, and Microsoft are watching this
- 7:34 The Amazon angle
- 8:20 Qwen 3.5 playing Doom on a smartwatch and what local AI means for builders
What the Anthropic lawsuit is about
The Pentagon keeps a blacklist of companies that can't bid on military AI contracts because of security concerns or ethical issues. Sometime in the last few months, Anthropic was added to it. Most companies would accept that and go back to commercial work. Anthropic didn't. They built their whole brand around AI safety and responsible development, so getting blacklisted for ethical reasons hits right at their identity.
The lawsuit was filed in the DC District Court on a Tuesday. Anthropic claims the blacklisting was arbitrary and violated its due process, and says the Pentagon didn't give proper notice or a chance to respond before adding them to the list.
Why does a blacklist matter so much? It's bigger than current contracts. It tells the whole defense world which companies are trusted. Once you're on it, defense contractors get hesitant to partner with you and government agencies may avoid your products entirely. In that world it does serious damage to your reputation.
The three arguments
From what's been reported on the filing, Anthropic makes three main arguments:
- The blacklisting process was procedurally unfair.
- Being blacklisted damages their reputation and business prospects unfairly.
- Excluding them makes military AI less safe, not more safe.
That third one is the bold part. Anthropic is basically saying the Pentagon needs them more than they need the Pentagon, because shutting out the company most focused on AI safety keeps the military from using the most advanced safety research.
I'm open about my bias here. I use Claude every day. My AI agents run my YouTube pipeline, newsletters, email strategy, and podcast outreach through OpenClaw running on Opus 4.6. So I think Anthropic has a strong hand, with a product the government probably needs.
Why the Pentagon might see it differently
There's another side to this. The Pentagon may have blacklisted Anthropic because they're so vocal about safety. If you're the military and you want companies that build what you ask for without pushing back, a company that might refuse certain projects or talk publicly about military AI capabilities could look like an unreliable partner.
Anthropic has been very public about its constitutional AI approach, has published research on alignment and safety, and has criticized other companies, OpenAI included, for moving too fast. Meanwhile the Trump administration has talked about speeding up AI development, especially for military and law enforcement use, with less emphasis on safety research and more on competitive advantage.
So the lawsuit forces a question: are AI safety and government contracts compatible or not? If Anthropic loses, it sends a signal that companies have to choose between ethical AI work and government contracts.
Why every AI company is watching
OpenAI, Google, and Microsoft are all watching from the sidelines. If Anthropic shows that safety research makes a company more valuable to the military, I expect every AI company to double down on safety work. If the Pentagon wins and can shut out companies that put safety ahead of capability, I think a lot of companies will scale back their ethical AI research.
There's also an angle not many people are talking about. Anthropic raised $4 billion from Amazon last year, and Amazon has major government contracts, including a $10 billion cloud contract with the Pentagon. Blacklisting Anthropic could create tension with one of the Pentagon's biggest cloud providers.
Qwen 3.5 playing Doom on a smartwatch
The second story: someone got Qwen 3.5, Alibaba's small language model, running on a smartwatch and playing Doom. This is the 0.8 version with 800 million parameters, which is tiny next to models like Claude and GPT that have hundreds of billions. It runs locally on the watch, understands the game state, and controls the character, deciding where to move, which enemies to shoot, and how to get through levels. I saw this in a video. I haven't tried Qwen 3.5 myself yet.
My plan is to test it on my OpenClaw setup for smaller cron jobs and easier tasks and compare it to Gemini 2.5 Flash, which handles those tasks for me now. The long-term goal of my mission control system is to run everything locally and avoid monthly API costs.
A year ago you needed a high-end desktop with an expensive GPU to run any AI locally. I think by the end of the year a lot of us will be running open-source models on something like an Nvidia DGX Spark, a Mac Studio, or our own machines, as long as hardware prices stay sane.
For builders, this opens up apps that run directly on your users' devices instead of depending on API calls to OpenAI or Anthropic. That means faster, more private, and no ongoing costs. If you want to start building, check out how to build an app with AI and my notes on AI agents.
Key takeaways
- Anthropic sued over being placed on a Pentagon blacklist for military AI contracts, claiming the process was arbitrary and violated due process.
- Its boldest argument is that excluding the most safety-focused company makes military AI less safe.
- The outcome could decide whether AI companies invest more or less in safety research when they want government work.
- Amazon's $4 billion investment in Anthropic and its Pentagon cloud contract add another layer to the dispute.
- Qwen 3.5 at 800 million parameters playing Doom on a watch shows how close we are to capable AI running locally on everyday devices.
FAQ
Why is Anthropic suing the US government?
Anthropic was added to a Pentagon blacklist that blocks companies from bidding on military AI contracts. It filed suit in the DC District Court, claiming the blacklisting was arbitrary, violated due process, and happened without proper notice or a chance to respond.
What are Anthropic's main arguments in the lawsuit?
From what's been reported, there are three: the process was procedurally unfair, the blacklisting unfairly damages its reputation and business, and excluding Anthropic makes military AI less safe.
Why would the Pentagon blacklist a safety-focused AI company?
My take is the Pentagon may see a company that speaks out about safety as an unreliable partner, one that might refuse certain projects or talk publicly about military AI capabilities.
Can a small AI model really run on a smartwatch?
Someone got Qwen 3.5's 0.8 version, with 800 million parameters, running locally on a smartwatch and playing Doom, reading the game state and controlling the character. I saw it in a video but haven't tested it myself.
Chapters
Transcript
Read the full transcript
0:00The company that makes Claude Anthropic, just got sued by the US government, and it isn't some small dispute over a contract. Anthropic is taking the Trump administration to federal court over being placed on a Pentagon blacklist for military AI work. This is insane. This is one of the biggest AI stories today, and no one is talking about it how massive this actually is. So, I'm going to break it down for you because this lawsuit is going to change everything about how AI companies work with the government. And at the end, I'll show you something that's completely insane. Someone got an AI model small enough to run on a smartwatch to play Doom. The local AI government is getting absolutely wild. But first, this Anthropic situation is crazy, makes my
0:46head spin. This is the first time a major AI company has actually been sued by the US government over military AI ethics. And it's happening right as the AI industry is trying to figure out where it stands with its defense contracts. So, here's what happened. The Pentagon maintains what is called the Defense Initiative Unit Blacklist. It's basically a list of companies that can't bid on military AI contracts because of security concerns or ethical issues. And sometime in the last few months, Anthropic got added to that list. Now, most companies would just accept this quietly. You get blacklisted, and you move on. You focus on commercial work, but Anthropic's different. We all know that. They built their entire brand
1:31around AI safety and responsible AI development. So, getting blacklisted by the Pentagon for ethical reasons is like getting told you're too dangerous when your whole thing is about being the safe option. The lawsuit was filed in the DC District Court on Tuesday. Anthropic is claiming that the blacklisting was arbitrary and violated its due process. They're saying the Pentagon didn't even give them proper notice or chance to respond before adding them to the list. But here's the thing that makes this thing really interesting is the lawsuit also argues that being on this blacklist actually hurts national security because it's preventing the military from using the most advanced AI safety research. So think about that argument for a second.
2:17Anthropic is basically saying, "You need us more than we need you." They're positioning Claude as essential for safe AI deployment in military applications. And honestly, they might be right. See, the Pentagon blacklist isn't just about current contracts. It signals to the entire defense ecosystem which companies are trusted and which ones are not. When you're blacklisted, it doesn't just block you from direct military work. It makes defense contractors hesitant to even want to partner with you. It makes government agencies avoid your products entirely. It's a reputation killer in the defense world. But Anthropic has something most companies don't. They have leverage. Claude is legitimately one of the best capable AI models
3:02available right now. And I use Claude every single day for my AI agents that I run my entire YouTube pipeline down, newsletters, email strategy, podcast outreach. The whole model writes my copy. It directs my life essentially at this point with my businesses and my goals and my dreams and my vision. It helps. So my whole content calendar is plug and play. Everything that I used to do manually is now been automated through Open Claude which is running on Opus 4.6. And if you're the Pentagon and you're trying to deploy AI safely in military applications, you probably want a company that's focused mostly on AI safety involved in that process. Which is why I use Claude every day. It's not only safe, but it also just gets So now
3:47that I've seen the full legal filing happen right before all of our eyes. What's the most been reported is Anthropic is making three main arguments. First, that the blacklisting process was procedurally unfair. Second, that being blacklisted damages their reputation and business prospects unfairly. And third, that excluding them actually makes military AI less safe, not more safe. So, that third argument is really bold. They're essentially arguing that the Pentagon is shooting itself in the foot by excluding the company that is most focused on AI safety from future military projects. But, here's where this gets more complicated. The Pentagon might have blacklisted Anthropic precisely because they are so vocal about AI safety. Think
4:32about it from the military's perspective, though. If you want AI companies that will build what you need without constantly questioning if it's ethical or safe, you don't want companies that might refuse certain projects or leak information about military AI capabilities because they think the public has the right to know. Which they do. Anthropic has been very public about their constitutional AI approach and their focus on avoiding harmful outputs. They've published research papers about AI and alignment and safety. They've been critical of other companies for moving too fast, OpenAI. From the Pentagon's point of view, that might make them unreliable partners for sensitive military projects. This creates a really interesting tension in the whole AI industry right now. It is getting a
5:17little bit awkward. On one hand, you have companies like Anthropic that are trying to build AI responsibly and safe. And then on the other hand, you have the government that needs AI capabilities quickly, like yesterday, and they might not want partners who slow down development with safety concerns. So, this lawsuit is going to force both sides to pick a position, either AI safety and government contracts are compatible or they're not. There's no middle ground. The timing of this lawsuit is also really significant. I won't say anything about that, but the Trump administration has been talking about accelerating AI development, especially for military and law enforcement applications. There's been less emphasis on safety research and more emphasis on competitive advantage. If Anthropic wins this lawsuit, it sets
6:04a precedent that AI companies can be blacklisted just for prioritizing safety research. Sounds weird. If they lose, it sends a signal that companies need to choose between ethical AI development and government contracts. But here's what I think is really happening here is the lawsuit isn't just about military contracts, it's about the future relationship between AI companies and government regulation. Anthropic is essentially betting that they can force the government to work with them rather than around them. It's just become a big ego play at this point. And I think they might be right because the alternative for the Pentagon is to work with companies that are less focused on safety. And if something goes wrong with military AI deployment, if there's an accident or a security breach, the
6:49Pentagon is going to want to be able to say they worked with the most safety-focused companies available, which isn't true. This is also sending shockwaves through the rest of the AI industry. OpenAI, Google, Microsoft, they're all watching very carefully from the sidelines because how is this lawsuit plays out is actually going to determine how all of them approach government contracts, safety research, all going forward. So if Anthropic can successfully argue that safety research makes them more valuable to the military, not less valuable, then every AI company is going to double down on safety work. But if the Pentagon wins and can basically exclude any company that prioritizes safety over capability, then you're going to see a lot of companies scale back their ethical AI
7:34research. Not good. Now there's another angle to this that no one's really talking about is Anthropic raised $4 billion from Amazon last year. Amazon has a massive government contract including $10 billion cloud contract with the Pentagon. So, by blacklisting Anthropic, the Pentagon might be creating tension with one of their biggest cloud providers. I truly believe that this lawsuit is going to reshape how AI companies think about government work. And honestly, I think Anthropic is in a strong position here. I Most of my friends use Anthropic. They all use Claude. I don't know too many people using ChatGPT. I'm sorry. But, Anthropic is at the forefront of the AI space, in my opinion. And they have one of the best safety researchers. So, they have backing from
8:20Amazon, one of the biggest companies in the world. And they have a product that the government probably needs. So, let me shift gears for a minute and back up because there's another AI story that's absolutely blowing my mind. Someone got Qwen 3.5, a Chinese model, to run on a smartwatch to play the game Doom. This is insane. So, Qwen is Alibaba's latest small language model, the 0.8 version. It has 800 million parameters, which makes it pretty tiny compared to other models like Claude and GPT. They have hundreds of billions of parameters. But, apparently it's small enough and efficient enough to run locally on a smartwatch. And not just run, but someone actually got it to understand Doom's game state and control the character. The AI is literally playing
9:06Doom on a watch. I saw this in a video, and it's blowing my mind. The model is making decisions about where to move, which enemies to shoot, and how to navigate through the levels. All of this running locally on a device that fits on your wrist. It's pretty crazy. This is exactly what I've been talking about with local AI, and I do believe that a lot of us, by the end of the year, are going to be running a lot of these open-source models locally on something like a heavy-duty DGX Spark from Nvidia, or like a Mac Studio, or maybe you build out your own machine. Barring that, hardware doesn't go to a million dollars for a stick of RAM. And I think that's going to help us keep more efficient while staying capable. It's, you know, like a year ago you needed very high-end desktop with an expensive GPU just to run any AI
9:53locally. Now that these open-source models are getting better, does Claude or OpenAI even fit into our equation as small, you know, founders, startups, and people who are just getting going? Probably not. So, we're talking about running capable models on watches. That's insane. I haven't tried Qwen 3.5 myself, but I'm definitely going to try it on my open class setup just for like smaller cron jobs to be able to handle some of the easier tasks and just see how it compares to the Gemini 2.5 flash that we currently have handling those tasks. So, the whole goal of my mission control system is to run everything locally down the road and avoid these monthly API costs. But, the thing with the Doom demo is not just a cool trick. It just really shows that we're approaching a world where any
10:40device can have AI capabilities built in. Your watch, your phone, we've already seen that. Your car, your home devices, all running AI locally without needing to connect to the cloud. And for builders like us, this opens up completely new possibilities. Instead of building apps that depend on API calls to OpenAI or Anthropic, you can build apps that run directly on your own users' device. Faster, more private, no ongoing costs. I think we're going to see a massive shift towards local AI by the end of this year. The models are getting too good and they're getting better and better. They're small enough to run everywhere and companies are realizing that depending on external AI API costs, it creates security and cost problems. The combination of these two stories today really shows you where AI is headed. On one side, you have big
11:26companies fighting over military contracts and government relationships. On On other side, you have AI becoming accessible to the point where it runs and plays video games on a smartwatch. Both trends matter for anyone building with AI right now. The Anthropic lawsuit is going to determine how AI companies work with governments and how much safety research gets prioritized. The local AI stuff is going to determine how accessible AI becomes and whether you need to pay a monthly fee to use it. If you're building AI products, you need to be watching both of these stories and trends. The regula- the regulation environment is changing fast and the technical capabilities are advancing even faster. If you want to stay of the head of this stuff, and I know it's overwhelming, especially if you're just getting into this space, and you want to
12:11actually learn how to build your first AI app and use these tools to build something that brings value, then you can check out link down below. We have a community of 32 members over the last few days inside of Shipping School, where we not only just talk about it, we actually build stuff. We work with you hand in hand to launch app, help you integrate AI in your current business workflow. We have an open clock course and a cloud code course that help you get you off the ground. So, whatever you uh latest tools that come out, we're going to be talking about it, we're going to be sharing with it. I even have a Next.js starter kit to help you get your web app off the ground. We do three boot camps a week with a fourth one on the way, where we basically walk you step by step through the process. Really, the whole goal is to go from
12:57idea to live product as quickly as possible. Because the AI landscape is completely changing, it seems like every day there's something new, and it's happening so fast that you don't want to wait too long because the opportunity will pass you by. They're not around forever. The Anthropic lawsuit is going to play out over the next few months, maybe a couple years, but the local AI is happening right now. The tools are available right now, and the opportunities are there right now for anyone willing to learn and ship fast. You got to start dirty, you got to start before you're ready, and it's time to get off the sidelines and start playing. So, whether you're interested in building with local models like Quinn, or you want to understand how to navigate the changing regulatory environment, or you just want to learn how to ship AI products quickly, that's what Shipping School is about. That community just launched a few days, and
13:43we have already over 30 members. Links in the description, and if you guys haven't subscribed already, please do so because we're releasing updates every single hour. We'll see you in the next one. Bye-bye.
Ready to start building with AI?
Join the Shipping Skool community free. Paid members get live bootcamp calls every week and weekly-updated courses.
Join Shipping SkoolKeep reading
Claude Duet Explained: Features, Safety, and How It Compares
Claude Duet is Anthropic's new proactive AI agent. It runs on Opus 5.5, works on its own cloud computer, connects to 6,000+ apps, and waits for your approval before it sends, pays, signs or deletes anything. It's a waitlisted research preview for US Claude Pro and Max subscribers.
Claude Mythos Preview Explained: Anthropic's Project Glasswing
Claude Mythos Preview is an unreleased Anthropic model that finds and exploits software vulnerabilities. Anthropic isn't selling it. Under Project Glasswing, it's working with companies like AWS, Apple, Google, and Microsoft to use the model for defense and get those security holes patched.
Claude Just Doubled Its Paying Users. Here Is What That Means for Builders.
Anthropic'''s Claude paid subscriptions more than doubled in 2026. An IPO at $60B+ is on the table. Here is why builders who bet on the Claude ecosystem are in the strongest position right now.
