What Anthropic’s White House Meeting Really Means - DTNS Weekend

0:00Business is booming, but you know what really helped? Zero. The accounting platform for small businesses. I used to think spreadsheets were just part of it. Ugh. But real-time cash flow? Wow. Invoices that chase themselves? Even better. Books that automate? That's a big one. And I can keep track from my phone while I get the job done. No spreadsheet in sight. What a life. Do right by yourself and your business and search Xero for your free trial. Conditions apply. Have you ever been stuck on a weight loss plateau, trying everything and anything you can to lose that extra weight and reach peak health?

0:37We've all been there, but Noom's unlocked the secret to reaching the mountaintop. Go in micro. The Noom GLP-1 microdose program starts at$99 and is delivered to your door in seven days. Start your microdose GLP-1 journey today at Noom.com. That's N-O-O-M dot com. Noom. Microchanges. Big results. Noom GLP-1-RX program involves healthy diet, exercise, and support. Individual results may vary. Meds and personalization based on clinical need. Not reviewed by FDA for safety, efficacy, or quality. No affiliation with Novo Nordisk, Inc., the only U.S. source of FDA-approved semaglutide. Not available in all 50 U.S.

1:14states. Parles-tu français? Hablas espanol? Parli italiano? If you've used Babbel, you would. Babbel's conversation-based technique teaches you useful words and phrases to get you speaking quickly about the things you actually talk about in the real world. With lessons handcrafted by over 200 language experts and voiced by real native speakers, Babbel is like having a private tutor in your pocket. Start speaking with Babbel today. Get up to 55 % off your Babbel subscription right now at babbel.com slash acast. spelled B-A-B-B-E-O dot com slash ACAST. Rules and restrictions may apply. My perfect day as sand, salt, water, and friends.

1:57But my moderate to severe plaque psoriasis can take me out of the moment. Now I'm all in with clearer skin thanks to Sky Rizzy. Riz and Kizum at RZA, a prescription-only 150mg injection for adults who are candidates for systemic or phototherapy. With Sky Rizzy, most people saw 90 % clearer skin and many were even 100 % plaque-free at four months. SkyRizzy is just four doses a year after two starter doses. Don't use if allergic to SkyRizzy. Serious allergic reactions, increased infections, or lower ability to fight them may occur. Before treatment, get checked for infections and tuberculosis. Tell your doctor about any flu-like symptoms or vaccines.

2:34Thanks to SkyRizzy, there's nothing on my skin. And that means everything. Nothing is everything. Ask your doctor about SkyRizzy, the number one dermatologist prescribed biologic in psoriasis. Visit SkyRizzy.com or call 1-866-SKY-RIZZI to learn more.

3:03Hello, my friends. Welcome to a weekend edition of Daily Tech News Show, where we're going to take a little closer look at a topic. by now a lot of what we're going to talk about may be in a little bit clearer focus but i wanted to talk about the broader outlines of anthropic walking into the west wing on friday to say hey white house wouldn't you like to be safer from all of these horrible things out here well if you want mythos you got to talk to us and when i saw a story that was about ai companies the U.S. government and a fight over policy. I had to bring on Justin Robert Young from Politics, Politics, Politics.

3:41How it's going, buddy? It is going fantastic, Tom. It is always a treat to talk to you and the DTNS listenership. I have decided when I make a malapropism like how it's going, I'm just going to use that as proof I'm human. I feel like how it's going was a short run series on ABC in 1978. How it's going. How it's going. Starring. Yeah. Damon. Hey, so how it's going here is that Anthropic CEO Dario Amode is getting to talk to what I understand from listening to Politics, Politics, Politics, one of those powerful people in the United States, White House Chief of Staff Susie Wiles. Seems like if you want the Trump administration to do something, that's who you talk to.

4:27And of course, the ground that he is trying to deal with here is that Anthropic is in two lawsuits with the U.S. government over whether they should be declared a supply chain risk or not. There has been no negotiation over that to speak of since those two lawsuits were filed. and what Anthropic would like is for it not only not to be declared a supply chain risk, but maybe do some business with the US government again on its own terms. How significant is this? It is very significant, but probably not on the level that a lot of people are looking at. The number one thing that is at play here is Anthropic's race to an IPO this year.

5:16they are in a race with open ai to do so there are hundreds of millions if not billions if not trillions of dollars at stake for these companies what you've seen over the past week and a half is a realization that overproduction of compute seems less of an issue today than it did a few months ago and now you look at those gigantic capex numbers for companies that are trying to build data centers and you can understand, no, no, no, this is a real jump for token usage. I say all that to say this. Anthropics in a particularly precarious place because they've had a real, it's the best of times, worst of times, 2026 right now.

6:02They have exploded in popularity. They became cause celebrity after they fought with the department of war. They surged their consumer usage. And yet now they found themselves hitting themselves at the top of their compute ceiling. And there's not a whole lot of ways that you can get out of that. You can buy more expensive compute, which they're going to have to do, but they didn't buy, they didn't invest in data centers a year ago when open AI was doing it. That is very important when it comes to the Pentagon for this reason. They were early on this. They were first on this. They were through various different partners in the Biden administration, given the ability to be a top DOD partner, CIA partner, very, very early on.

6:49When they got into that spat with the Department of War and specifically some other people that are also Silicon Valley adjacent inside of it, they did not send a great signal to either a their funders and the people that are really invested in this ipo or other big enterprise people for whom now you set a precedent that okay if you're going to tell the government to pound sand on something that you will not assuage them that you won't turn off their api if you believe that something is not to your liking then what do you say to everybody else? Now they always had the best coding model, although that's more in debate now, but let's just go ahead and assume that that's the case.

7:38If you have already complicated your world with your surge in consumer demand, and you don't have the money coming in for this government work, and you're compute constrained, you have to solve one of those problems. You want to solve all those problems before the IPO. So what they did, and they being anthropic, is they hired Ballard Partners. It's a big lobbying firm within D.C. that has ties to the Trump administration. What have they done over the last week? They sent Dario to D.C. He talked with a few right wing think tanks, including Heritage. Those talks didn't go well, apparently. The words have leaked out that Dario might not be the face to make peace with some of these organizations.

8:27But still, the 800-pound gorilla is big chungus, and he sits in 1600 Pennsylvania Avenue. What I suspect is happening right now is that Dario is talking to Susie Wiles to talk about what this deal is going to be. anthropic knows they need to make a deal and this is what i suspect ballard partners again a very good lobbying firm that knows all the right people to talk to and how to frame a story the story that comes out in axios is to me the golden bridge where everybody gets what they want mythos despite the fact that it seems to be a little bit of these steamed hams of models uh It has a lot of tremendous claims, but nobody's allowed to see it.

9:15That remains the big, potent, oh, my Lord, it could end the world, amazing achievement for Anthropic. And so Dario and Anthropic remain this state-of-the-art amazing frontier lab. And the government needs it. But on the other side, and this is why I don't think you're going to hear a ton coming out of this unless it goes really, really badly, which it always might. you're not going to hear a lot more because what I think the other side of this is that Dario just takes the deal that Sam Altman took with the department of war. When all of this went down, the same one that Dario called an internal Slack conversations, a capitulation.

9:55And, uh, which is one of the reasons the supply chain risk thing went on the table is that that Slack message got out and the U S got mad. Uh, well, that was our, that was in the works before. It certainly didn't help any of the last minute peacekeeping that was trying to help them. It helped turn what was a threat to to come to heel into a threat to bash you in the head. The larger problem was that Dario and the Department of War didn't get along. And I think that now you're looking at a situation where. The D.O.W. doesn't want to change their models. Palantir doesn't want to change their models.

10:30They might be working in other stuff. Well, and that's an important point to point out to folks is the U.S. military is still using Anthropic right now. It's in a six-month grace period. So it's not like they shut off Anthropic the day they declared them a supply chain risk. They said, we're going to wind down the use of you over the next six months. No, they then went and used it in Iran. Yeah. So this is currently being used. Nobody wants to change anything that is working. And it appears that Department of War and Palantir, which is the main contractor that's working with the Department of War, Anthropic is a subcontractor of Palantir that they would like to keep.

11:08they would like to have the option and i think anybody who's at the state of the art right now for business or government you want access to all the top models and there's no doubt that mythos or or anything else is going to be something that will be among the top models as much as i like to joke about the fact that mythos is the girlfriend in canada there is i don't have any doubt that it is an exceptional model i think it is it is tremendous this i mean they are they're very very good at what they do but you want to have access to everything you want to have access to whatever open ai is cooking up with spud you want to have access to mythos you want to have access to i mean when you're talking about that kind of stuff but anthropic needs a golden bridge and when i read that story in axios today that's what i saw gleaming yellow steps that lets everybody get what they want dario can quietly take the open ai deal mythos remains this vaunted thing and Anthropic has a higher esteem and everybody can move along.

12:32done. No spreadsheet in sight. What a life. Do right by yourself and your business and search XERO for your free trial. Conditions apply. Breaking news for your home. Wayday. Wayfair's big sale is almost here and you can score the best deals in home. Like up to 80 % off everything home. Yes, you heard that right. 80 % off with free shipping on everything. Plus surprise flash deals all way day long. We're talking outdoor furniture and essentials, rugs and more. All at Wayfair. And all up to 80 % off during Wayday. Shop Wayfair's Wayday from April 25th through 27th. Wayfair, every style, every home.

13:12Have you ever been stuck on a weight loss plateau? Trying everything and anything you can to lose that extra weight and reach peak health? We've all been there. But Noom's unlocked the secret to reaching the mountaintop. Go in micro. The Noom GLP-1 microdose program starts at$99 and is delivered to your door in seven days. Start your microdose GLP-1 journey today at Noom.com. That's N-O-O-M dot com. Noom. Microchanges. Big Results. Noom GLP-1-RX Program involves healthy diet, exercise and support. Individual results may vary. Meds and personalization based on clinical need. Not reviewed by FDA for safety, efficacy or quality.

13:50No affiliation with Novo Nordisk Inc., the only U.S. source of FDA-approved semaglutide. Not available in all 50 U.S. states.

14:00If you've used Babbel, you would. Babbel's conversation-based technique teaches you useful words and phrases to get you speaking quickly about the things you actually talk about in the real world. With lessons handcrafted by over 200 language experts and voiced by real native speakers, Babbel is like having a private tutor in your pocket. Start speaking with Babbel today. Get up to 55 % off your Babbel subscription right now at babbel.com slash acast. Spelled B-A-B-B-E-L dot com slash acast. Rules and restrictions may apply. Security program on spreadsheets, new regulations piling up, and audit dread?

14:40It's time for Vanta. Vanta automates security and compliance, brings evidence into one place, and cuts audit prep by 82%. Less manual work, clearer visibility, faster deals. Zero chaos. Call it compliance, or call it compliance. Get it? Join the 15 ,000 companies using Vanta to prove trust. Go to Vanta.com slash com.

15:08Let's talk a little bit. I want to talk about the OpenAI deal, then I want to talk about Mythos. The OpenAI deal is Sam Altman signed a deal that said, we will still be able to object if there is a misuse of our models that we don't like. And the two main things are, are you going to have automated systems where they're actually able to fire up with lethal force in an automated fashion? That is somewhat, as I understand it, a theoretical issue. It's not something that they can or are going to do today. It's something they would like to explore and test. Getting into gigantic world-changing fights about theoretical issues is literally the origin story of Anthropic for the record.

15:58Yeah, and I think that's one thing that gets lost is some people I've talked to are like, yeah, well, I don't want robots to just be going and firing. What if they malfunction? I'm like, that's not really what this is about. This is about the idea of using an Anthropic model to test the idea of an autonomous system at some point in the future. So this is very theoretical. And like you say, Anthropic does like to act as if theoretical. It's the reason why they exist. The people that came out of OpenAI to start Anthropic started because of hypothetical fights like this. And so I think what OpenAI said is like it's kind of theoretical and, you know, we don't have to put it in the agreement for them.

16:41We can restrict their API. There are other parts of the deal we could use to just slow things down. and there's no need to pick a fight and put a strict thing in there the way Anthropic is doing because Anthropic is more idealistic than OpenAI and these sorts of things. OpenAI tends to try to be more pragmatic or appear more pragmatic. Anything else that you want to add to that? Because I just want to make sure people understand that it's not that I'm saying autonomous robots firing are fine. It's that I think we're counting angels on the head of a pin when you have that fight. it's an important conversation yeah what i always think from the very beginning of this fight and this is from all angles the open ai angle the anthropic angle and the department of war angle the biggest sin from all of them is that anything spilled out into the public yeah these are not the kinds of issues that ever see the light of day every once in a while you'll hear about a fight from like, you know, Northrop Grumman or Boeing or something like that, that'll come out in the press years after the arguments had happened.

17:52Like it'll be about some weapons system that failed or some plane that failed. And then there'll be a big story in the Wall Street Journal. And there'll be anecdotes about how, oh, the CEO was screaming at the Secretary of Defense and they hung up the phone or some threat was exchanged. But these are like ancient history by the time that we get to know it. We were live blogging the things that were being said between Pete Hegseth and Dario Amodi. That was the problem. The problem was from the Department of War's perspective that they were afraid they were going to be in the middle of kinetic action and an API was going to get restricted.

18:30And from Anthropik's point of view, that they felt contrary to their agreement, something would be done that they wouldn't like. Now, I tend to believe that that line for anthropic is sometimes even theoretical, that they will pull their support. They will make a stink about something that is steps away from happening. But again, that's their nature. There had to be a meeting of the minds. And I think that that's kind of what's going to happen is that there's going to be a more realistic conversation about exactly what's going to be used and probably things that we won't know about. The Department of War said with the autonomous robots that they wanted Anthropic on the board that they're putting together to make these things as safe as possible.

19:18Yeah. And to evaluate the policies that would go around them and decide when they could be used, if they could be used and all of that. But by the time that it becomes a public thing. Now you're at a political conversation of who do you trust? Do you trust Dario? Do you trust Pete Hegseth? And everybody has preconceived notions about all these people. Yeah. And the narrative is then created by what your political leaning is. And you can create it whatever way you want. The fact of the matter is they got in a shouting match over something that was resolvable, not something that was like an imminent threat.

19:54The other one was mass domestic surveillance, but it was of unclassified commercial bulk data. It was Anthropic objecting to the US government doing what advertising companies do all the time, which is buy data from a data broker and then use Claude to analyze it. I perfectly respect Anthropic's right to say, yeah, we don't want you to use Claude for that. I think that is also eminently resolvable, especially because this contract is with Palantir because you could just say, well, Palantir will use a different model for that stuff because Claude was not being used primarily for that. I mean, again, it's so hard to make definitive statements when we're talking about relationships that we don't know and are deliberately kept secret.

20:42Business is booming. But you know what really helped? Zero, the accounting platform for small businesses. I used to think unpaid bills were just part of it. Ugh. But batch payments? Wow. Automated accounts payable? Even better. All in one place? That's a big one. And I can keep track from my phone while I get the job done. No bills in sight. What a life. Do right by yourself and your business and search XERO for your free trial. Conditions apply. Legendary doesn't happen by chance. It happens because it's what BetMGM does. Sign up for BetMGM and receive up to$1 ,500 back in bonus bets if you don't win your first bet.

21:22Get into hockey, basketball, and more of your favorite sports with same-game parlays, live betting, and more. BetMGM. Make it legendary. BetMGM and GameSense remind you to play responsibly. See BetMGM.com for terms. 21 plus only. This specific promotional offer is not available in Michigan, Mississippi, New Jersey, New York, Nevada, Ontario, Pennsylvania, Puerto Rico, or West Virginia. Gambling problem? Call 1-800-GAMBLER or 1-800-MY-RESET. Available in the U.S. For New York, call 877-8-HOPE-NY or text HOPE-NY 467-369. For Massachusetts, 1-800-327-5050. For Iowa, 1-800-BETS-OFF. For Puerto Rico, 1-800-981-0023.

22:01First bet offer for new customers only, subject to eligibility requirements. Rewards are non-withdrawable bonus bets that expire in seven days. In partnership with Kansas Crossing Casino and Hotel.

22:14the department of war denies that this kind of stuff was was in any way even really controversial that claude is almost assuredly used for very similar things by lots of other companies and by lots of other companies so uh you know i don't know what in particular they were objecting to it certainly was something that the department of war pushed back against but again a lot of this apparently the stuff that the department of war was really worried about was palantir or sorry anthropic was demanding information about how its models were used with the venezuela raid for example and the department of war looked at that as a complete non-starter that a subcontractor for a contractor would be demanding from the department of war top secret military information.

23:11That's really what kind of kicked this thing off into a bad direction, at least according to all all reportings. And I think that cooler heads have prevailed. Everybody's realized, look, we're better together. And to Anthropik's credit, Anthropik very much believes that the United States should be the leader in AI for all of the effective altruism things that tends to create a lot of these hypothetical freaks at freak outs. One of the things that they are dead set on is that China cannot win the AI race and that the United States of America and specifically the government should have access to the top tier tech.

23:47So coming together on that is, is, you know, a process, but it looks like we are moving toward a reconciliation, which is always the smartest way that this could have ended. Yeah. So, so there's lots of ways Anthropic can declare victory here and say, you know, if they're put on the policy board, we will get to help decide how these autonomous things are used and something around the surveillance data, et cetera, et cetera. Let's talk about the stick, which is the idea that Anthropic released mythos in a limited manner because it is so good at finding vulnerabilities. It would be irresponsible to let just anyone have it.

24:29And gosh, among those people should be the U.S. government. You would like to use mythos, right? Well, if you're calling us a supply chain risk, you can't. Let's talk. How much of that do you think is real and how much of that is a good excuse for Susie Wiles to say, OK, yeah, we can use that as the story. Let's talk. my assumption is that this is the golden brick this is actually the carrot for anthropic that anthropic gets to say we we won on something because we created the best stuff and we were able to demo something for suzy wiles that blew her hair back we were able to demo something for the department of war that blew their hair back and they said that they needed this and we have created an agreement that we are comfortable with and they can then take that now just a little commentary and again i produce the open ai podcast consider me a shill if you would like i have some questions as to exactly how a massive model like mythos which by all accounts this is a behemoth of a model is going to be served at scale when anthropic is having a hard time keeping their servers online right now for the current offerings that they have, which has created a lot of consternation amongst power users because there is commentary out there about the effectiveness of their models as they deal with that strain.

25:57That being said, I have no doubt that Mythos is amazing. I have no doubt that Mythos is top of the line. I believe that it belongs with the government. This was always a problem that was solvable, that just needed cooler heads. I tend to believe that yeah of course the department of war wants it of course the american government wants access to the the most top tier models and again this is the ai industry in the year of our lord 2026 state of the art lasts you know uh for for about the length of mcdonald's french fries like we're we're at a pace where this stuff is changing so rapidly that yes it makes sense right now, when the air around Mythos is so, dare I say, mythological, that you would want to leverage it to get the deal.

26:51I do think ultimately it's going to have to result in Dario understanding that the deal that he is going to have to agree to is very close to the one that OpenAI did. Yeah. I'll say the following things can all be true. Mythos can be the most competent model today at finding vulnerabilities. Mythos could be misused by people to find vulnerabilities. Mythos could be way too expensive for Anthropic to run at scale for its entire user base. And so a neat little package is to say, too dangerous. That solves all of our problems. Another company may come out with an equivalent model and make a different decision because it has different compute capacity and different tolerances of risk and different ways of guarding against that risk.

27:43But what Anthropic did is perfectly reasonable for Anthropic. Very, very on brand for them. I will. And also it really does. I mean, I, I, I say that without snark, I think Anthropic is more cautious on things. And so even if they didn't have the compute restraints, they might've made the same decision anyway, but it certainly is, you know, a nice side effect that they're like, well, and it saves on compute as well. I, yeah, they have not said that for the record. That is my supposition that, that, uh, I also share that supposition of like, I just, just from the outside, I don't think it's controversial to say like, you guys are running against token limits.

28:20Uh, you're having to change your pricing. Uh, clearly anything that saves you compute is going to be a good thing for you. I don't, I, I don't think they could reasonably look at, look at, look at that uptime chart, the status chart for, for, for Claude. And when it looks like an autumn day with all of those reds, oranges, and yellows, then it is not a good time for them. Yeah, I do think that they've got real compute issues. And it is something that OpenAI has not been shy about leaning on the neck of their rival to say they screwed up. We didn't. We have a higher compute ceiling. We are going to push things forward.

29:01I don't know anything proprietary. So let me just say things that are public. OpenAI came out on Wednesday with their cybersecurity framework, therefore pointing out exactly how serious they take cybersecurity, exactly how people can utilize their models to find vulnerabilities, the same vulnerabilities that Mythos was said to tackle. I also think next week, this is not any kind of state secret, but OpenAI has got a new model. There's a lot of buzz around it. They've got a new image model that's currently in the LM arena, or at least allegedly called duct tape. That is getting a lot of buzz. I'm going to presume that both of those will probably see the official light of day at some point in the next seven days.

29:46And then we'll have a conversation about where the state of the art is. And that's where I think right now, if I were Dario, I would make this deal right now. because I don't know exactly if Mythos is going to have the same gravitas to it next week. If OpenAI has a model that is a huge step change beyond 5.4, which is very, very well received, that they're going to have the same gravitas to them because OpenAI is going to make that model available for everybody. Well, Justin, thank you for helping us understand the pieces on the board as people hear what comes out of all of this. It really does, I think, provide a service to people to say like, oh, OK, I would think this if I saw that headline.

30:31But knowing the players and the pieces and the motivations helps me interpret it at a little deeper level. Really appreciate that. Thank you very much. If folks want to follow this kind of analysis for more than just technology, where should they go? They should go to politics, politics, politics dot com. That's where you get my twice a week free show. And if you subscribe on Substack, you get another two episodes four times a week. Me yammering in your ear about the political world. The the one you get as a subscriber about the Sunday, Sunday, Sunday shows is must listen for me to understand what is coming in the next week.

31:07Thank you, Justin Robert Young, as always, my friend. Thank you. And thank you, folks, for supporting us at Patreon dot com slash DTNS. We will be back on Monday with all fresh new news shows. Talk to you then.

31:27The DTNS family of podcasts. Helping each other understand. Diamond Club hopes you have enjoyed this program.

31:39Owning a home is full of surprises. Some wonderful, some not so much. And when something breaks, it can feel like the whole day unravels. That's why HomeServe exists. For as little as$4.99 a month, you'll always have someone to call. A trusted professional ready to help, bringing peace of mind to 4.5 million homeowners nationwide. For plans starting at just$4.99 a month, go to homeserve.com. That's homeserve.com. Not available everywhere. Most plans range between$4.99 to$1 ,199 a month your first year. Terms apply on covered repairs. My perfect day has sand, salt, water, and friends. But my moderate to severe plaque psoriasis can take me out of the moment.

32:17Now I'm all in with clearer skin thanks to SkyRizzy. Riz and Kizumav RZA, a prescription-only 150 milligram injection for adults who are candidates for systemic or phototherapy. With SkyRizzy, most people saw 90 % clearer skin and many were even 100 % plaque-free at four months. SkyRizzy is just four doses a year after two starter doses. Don't use if allergic to SkyRizzy. Serious allergic reactions, increased infections, or lower ability to fight them may occur. Before treatment, get checked for infections and tuberculosis. Tell your doctor about any flu-like symptoms or vaccines. Thanks to SkyRizzy, there's nothing on my skin.

32:53And that means everything.

32:59Ask your doctor about SkyRizzy, the number one dermatologist prescribed biologic in psoriasis. Visit SkyRizzy.com or call 1-866-SKYRizzy to learn more.