Everyday AI Podcast – An AI and ChatGPT Podcast - Ep 828: Anthropic Responds: Why Claude’s CEO didn't sign the open model pact and the real reasons why

Starting point is 00:00:00 This is the Everyday AI show, the Everyday Podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Anthropic finally responded as the last AI lab out in not supporting open source AI in the viral open weights and American AI leadership paper slash coalition. That's because Anthropics' CEO Darcy. Mario Amati just published a letter defending the company's decision to not join the other players. And I think it's actually more about Anthropics IPO than it is about open source. But here's the setup and the backdrop. InVIDIA and Microsoft put out a letter four days ago defending open models.

Starting point is 00:00:55 And basically everyone signed it. Meta, OpenAI, Google, IBM, everyone. And Anthropic was the only major tech player that said no. And late yesterday, Dario published Anthropics answer in a blog post, and it opens by insisting they've never wanted to ban open models, which is technically true in writing, but not really truly in line with their previous statements warning about how dangerous open models are and there should be government mandates for them. Now, here's the part that nobody's talking about. why Anthropic is really trying to get DC lawmakers scared of Chinese open models. It's one number, 80%. Yeah, that's roughly 80% of Anthropic revenues just comes from businesses paying per token.

Starting point is 00:01:45 While the other big players have a fraction of that. So free downloadable Chinese open source models literally choke Anthropics' only real source of revenue. And Anthropic filed for an IPO targeting in, an October launch. Hence the reason why Anthropic zigged, well, every other tech company zagged when it came to supporting open source. So today we're breaking now not just what the letter says, but where it actually even contradicts Dario's own previous words and why I think this was written for Washington policymakers and not for you and me. All right, let's get into it. Here is the big picture. Anthropic is the only major AI lab that refused to defend open models.

Starting point is 00:02:29 in the open weights and American AI leadership, a paper slash coalition that was put forth by Microsoft and Invita. Essentially, they said, well, open source is good for the American economy. It's good for national security. And we're going to talk a little bit more about that paper. But why? Well, Anthropic will tell you it's ultimately about safety and control, but I think it is that 80% number.

Starting point is 00:02:56 And we're going to look at, well, it might not work very well, if Anthropic can't get this essentially to go in their favor. And that's because their IPO is potentially only weeks away in October. So on today's show, stick with me and you're going to learn why that 80% meter token revenue makes Anthropic uniquely exposed to open models. You're going to know what the regulatory capture means when it comes to AI and why this letter just fits the pattern exactly. You're going to know why the timing points at Washington policy makers,

Starting point is 00:03:29 because there's an important date coming up in just a couple of days. And I don't think this letter was made for me and you. And you're going to know what Dario said in 2025 that flatly contradicts today's letter. All right. Let's get into it. Welcome to Everyday AI. If you knew here, my name's Jordan. We do this well every day.

Starting point is 00:03:47 It's your unedited, unscripted, live stream podcast and free daily newsletter helping everyday business leaders like you and me. Keep up with the nonstop AI avalanche. I tell you what's important. What's not? You use that decision and all of a sudden, wow, you're the smartest person in AI in your company. So it starts here, but please make sure to subscribe to the podcast and go to our website at your everyday AI.com. All right, let's get into it.

Starting point is 00:04:12 This has been all the talk on tech, Twitter and, you know, in the media, in anything AI, right? If you follow AI on a daily basis like I do, this conversation has been just grabbing every single headline. And it's extremely important and also a little bit nuanced. So today's show, I'm going to try to go through it a little bit quicker. Stick to just the facts. And my opinion, maybe I'll, you know, bring back an accidental hot take Tuesday for old time's sake. But make sure you go check out today's newsletter for more.

Starting point is 00:04:47 So here's what actually Dario put in his published response on open models. So the core claim was that Anthropic. has never advocated banning open-weight models. And he rejected the argument that open models actually help defenders more than attackers, which was a core claim in the Microsoft slash Nvidia paper that everyone could have got high. And Anthropics says it's better for chip controls to crack down on distillation and mandatory safety testing for all. And they deny that protecting Anthropics business ever motivated the position. That's a funny one.

Starting point is 00:05:26 All right. So let's take a quick look at the letter. I'm not going to read it, but I will kind of read the opening paragraph and go through some of the bullet points here. So this is from Dario Motti, Anthropics CEO. So yeah, this just came late last night, actually, or late afternoon. I'm so tired sometimes like 4 p.m. feels like late night. So it says over the last few days, there have been a lot of discussions about open weight models, especially those from China.

Starting point is 00:05:52 Reports suggest that some U.S. officials are considering ban. banning the use of Chinese open source models, sorry, Chinese open weights models by U.S. companies. In response, many tech companies have signed a letter supporting open weights models, and some people have even accused Anthropic of wanting to ban open source models as a means of protecting our business. Anyone who has read my past writing should know that I don't regard such bans as a useful measure, but let me state it clearly so that there is no doubt. Anthropic has never advocated for a ban on open weights models. Then he says open weights models that don't have dangerous capabilities are a public good.

Starting point is 00:06:32 They don't cost anything besides the compute needed to run them. And they provide value to businesses, developers, and researchers. All right. So let me just take a pause. And I probably should just kind of talk very briefly about open source, open weights models. So if you are brand new, maybe this is the first episode you've ever listened to and you're like, what the heck is going on? Why does this matter? well, a couple important points that Dario brought up here in the letter.

Starting point is 00:06:55 Number one, there has been this idea floated recently. And we'll see, you know, maybe in 10, 20 years, how much of Anthropics D.C. lobbying efforts maybe led to this, right? I'm a former journalist. I understand how this works. All the big labs have a lot of lobbyists. But the U.S. government is actually considering a ban on American companies using Chinese open source models for, you know, what they're saying, competitiveness, national security concerns,

Starting point is 00:07:24 etc. But essentially, open source or open weight models are models that you can, well, you can download and you can run them on your machine. Well, very small open source models or what open source models used to be, you know, like two years ago. Today's open weights Chinese models, you can't download those on your computer. You're literally going to need a spare data center to run the newest and the ones that kind of set this off. are Moonshots, Kimmy, K3, Z-A-I's, G-L-M-5-2. And I think that eventually we'll be talking a little bit more about Quinn 3.8. So essentially, you have all these models that, well, if you're a huge enterprise,

Starting point is 00:08:04 you can use them for free. But if you are using them through a hosting provider, through an inference provider, they're way cheaper, right? For the most part, not always, but usually these open source models are way cheaper. if you can't download and host them or run them locally. All right. So that's the big picture on what kind of Dario said in his letter in all of that. So are you still running in circles trying to figure out how to actually grow your business with AI?

Starting point is 00:08:40 Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on Gen AI. Hey, this is Jordan Wilson, host of this very podcast. Companies like Adobe, Microsoft, and Nvidia have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead. And some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use Gen AI. So whether you're looking for chat GPT training for thousands or just need help building your front end AI strategy, you can partner with us too, just like some of the biggest companies in the world do. Go to your everyday AI.com slash partner to get in contact with our team. Or you can just click on the partner section of our website.

Starting point is 00:09:28 We'll help you stop running in those AI circles and help get your team ahead and build a straight path to ROI on GenAI. Some of the more important bullet points here. He said, my primary concern is the risk that authoritarian governments, not solely the Chinese Communist Party or CCP, although the CCP is clearly the most capable threat. build AI models that are more powerful than those built in the U.S. and use them to achieve permanent military superiority or perpetrate incredibly deep repression of their own people. So essentially, he said the biggest concern is, well, the Chinese government is going to use these for military superiority. Number two, secondary concern, he says, is that powerful AI models may be misused to carry out cyber attacks or biological attacks, and those may have serious alignment issues. So, you know, I agree with both of those things, right?

Starting point is 00:10:27 I've been saying it all along. I think artificial intelligence, for the most part, it is going to be more important, maybe in five to ten years than, you know, what your traditional military might be looked at. It's going to be more important than oil or gold or anything else because the nation, essentially, that figures out artificial general intelligence and artificial superintelligence first has a legit unfair advantage on the rest of the world. and you can go from a mid-tier global power to the global superpower if you control artificial general intelligence or artificial super-intelligence.

Starting point is 00:11:02 So kind of to address these concerns, what he does say in his letter, he says, number one, we should not sell powerful chips or chip-making equipment to China, essentially saying, hey, we still have a lead, although it's maybe slimmer than it was before, but China can't catch us if they don't have our chips, right? Previously, they said, Anthropics said that by essentially creating a chip ban, the U.S. could get a 12 to 24 month lead in AI. I don't necessarily think that's true. The lead could extend from where it is now, which is like two months.

Starting point is 00:11:35 It used to be about six months between the U.S. frontier models and the Chinese open source models. So maybe it could extend it. I don't necessarily think it would be 12 to 24 months. That's a little bit of lofty thinking, I would say. And then Dario also says we should crack down on. industrial scale distillation distillation operations, right? So essentially how China is catching up, right? They do have some novel architecture and kind of ML approaches. But for the most part,

Starting point is 00:12:05 it's distillation, right? They're just copying, you know, millions of inputs and outputs from, you know, clawed models and other models to make their models smart. So they're just copying someone else's answers at a fraction of the cost. And then, Dario says, all sufficient capable models opening closed should go through mandatory safety testing. All right. There's more on the letter. It's not super long. I'm not going to read it all.

Starting point is 00:12:30 We'll make sure to put it in today's newsletter as well. So this is all essentially in response to the original kind of movement that Nvidia and Microsoft put forth on July 24th, which was this past Friday. And this was called the Open Waits in American AI. leadership. All right. So essentially, they circulated this arguing that open models cut costs and avoid vendor locked in. So they said it's good for American businesses. It's good for competitiveness. And then they also said it's better for safety because they said that the core safety claim of having open models is that defenders, right, because this whole thing comes down

Starting point is 00:13:12 to cyber attacks versus defending your business, right? One of the big arguments is, oh my gosh, if you have these, you know, these open source models everywhere, you can't keep track of them, right? And eventually, and I agree, probably in two years, I've said this, you'll be able to, you know, have a version or an open source model that is of the GPD 56 sole Fable 5 tier of models. You'll be able to have something that powerful on a, you know, consumer-ish desktop device. So at that point, you know, yeah, there is going to be, you know, critical infrastructure at risk from cyber attacks. You know, think of, you know, fishing scams and, you know, people holding, you know, web systems for hostage times a thousand and compacted into a very tight time frame,

Starting point is 00:13:57 right, once that does happen. All right. So you're thinking of hospitals, school systems, power grids, right? Entire small nations, you know, city and state governments, right? All these places are going to be extremely at risk. And I agree with that. But essentially, Microsoft and NVIDIA are saying, well, the best way to defend against this is, well, open models, because you of more people contributing. People are going to find these vulnerabilities, and it collectively makes, you know, everyone stronger. So it gives everyone better defensive tools, all right?

Starting point is 00:14:28 And essentially, everyone literally agreed with that from Amazon, OpenAI, meta, Google, IBM, everyone, except Anthropic. So my take on this is Dario's letter is not for us. This is literally for regulatory capture. This is to influence the policymaking in, DC. So there's been a lot of articles about, you know, just how big of an impact that Anthropic has in lobbying. And yeah, I think that this has to be the real reason here. But I mean, my take on the actual letter, it's well written. There's great arguments in there.

Starting point is 00:15:05 And it genuinely engages with real risk, right? There's nothing in there that I'm looking at. I'm like, this is absolutely blasphemy, right? Aside from the fact that they say that they absolutely never, you know, called for a ban on open models. They have just without using the word ban, right? I don't know. If you say something, you know, walks like a duck, quacks like a duck, is yellow like a duck, you know, is in the jersey on mighty ducks, but you're like, well, it's not a duck. Well, that's essentially what you said.

Starting point is 00:15:38 Just because you didn't say that animal right there is a duck, but you described every single duck characteristic and used the word duck in describing the characteristics, but then you're like, well, it's not a duck. Well, yeah, it is. All right. But the important thing here is Dario's letter comes just days before an important federal deadline. And that is why the regulatory piece is very important.

Starting point is 00:16:00 So President Trump signed essentially this AI order on June 2nd with a 60-day deadline. And guess what? That 60-day deadline is, well, this week. So the framework right now does not cover open weight models, right? So that's why there's all this kind of jockeying right now. And, you know, why you've had these stories come out in the past seven to 10 days, right, saying that the U.S. is now maybe considering kind of throwing open source models, not in this executive order, but potentially just banning them.

Starting point is 00:16:36 Because out of nowhere, right, you had, you know, moonshots Kimmy K3, I think, especially come up and, you know, essentially it is in that top tier. Whereas before, the open source Chinese models were never in that tier. So that obviously takes away from American business, number one, right? Because whether you know it or not, and I've been saying this all along, AI in large language models, right now they run the American economy. So when you have a Chinese open weights contender come in and well, potentially, you don't want to take all that spend, right, from, you know, putting America first sort of thing, which is the kind of the, kind of the trend. Trump's, the Trump people's position on this, right? Well, that takes away probably a lot of money that American businesses would be spending with other American businesses. So not only that,

Starting point is 00:17:28 but, you know, kind of the other chain of thought, ha, AI joke accidentally made, is that, you know, the more powerful these models become, well, they're just going to become better at distilling versions of themselves from the other frontier labs. And, you know, it seems like anthropic, maybe more. so than others seemingly is struggling. Maybe they're just the ones complaining a little bit more because they technically have more to lose. But the Dario letter lands just five days before this kind of order gets locked in. So the order is voluntary and there's no licensing and no preclearance requirements. But essentially, you're supposed to clear models with the government 30 days before

Starting point is 00:18:09 release to give everyone fair amount of time. So yeah, there's that. Pretty big thing that no one is talking about. But the real reason is the money, y'all. So these are according to public estimates, which are pretty close. A lot of it is from semi-analysis, which is kind of the leader in the space in terms of looking at compute and its impact on the economy, the money in and the money out. So essentially, the best estimates are 80% of Anthropics revenue comes from just selling tokens, right? Whereas its closest competitor, Open AI, it's estimated that it's only 20%. And then, you know, your other big tech companies, it's very small considered, right?

Starting point is 00:18:54 Because they have, you know, dozens or hundreds of lines of revenue like, you know, Google, it's only about 2%, Microsoft, 2%, Amazon 1%, and meta, not even 1% yet. So essentially, Anthropics' entire business model is just selling tokens, right? And well, what if now all of a sudden, if you just stop buying those tokens, if your company stops buying them, right? If you're like, hey, we don't need, right, we can get 95% of our AI use with a model like, you know, GLM 52. And that one is a little bit smaller. And well, you know, we can invest, you know, six figures and getting this set up. We were spending six figures or seven figures and AI each month.

Starting point is 00:19:40 Right. Now all of a sudden we cut our cost by 95%. We have more control. We're running it locally, et cetera. So you can see how this is in extreme. I am talking like potentially lights out for Anthropics growth, right? Not for them as a company, but for their staggering growth that we've been reading about through 2024 and 2025. And, you know, they've been the main beneficiary of this token maxing.

Starting point is 00:20:07 But now that token maxing is going away. And it's going toward token efficiency. So not only, right, if you look at open router, which kind of is a third-party tracker of, you know, who's spending tokens where, you know, it used to be Anthropic. Anthropic used to dominate. Now I checked yesterday. I think they were like number seven. They're like the seventh most used lab in open routers.

Starting point is 00:20:29 And they used to dominate, right? So people are coming to a couple of conclusions. Number one, yes, Anthropics still has probably the best models in the world, right? the combination of Opus 5 that just came out in Fable 5, although I do think GPD 56 sole is right there. And on many important benchmarks, I do think GPD 56 soul is still better. But regardless, Anthropic has some of the best models in the world. But 80% of their revenue is just selling tokens. And now all of a sudden, over the past 10 days, you have Kimmy K3 and we're going to be talking more about Quinn 3Aid and probably whenever GLM 53 comes out, all of a sudden, yeah,

Starting point is 00:21:07 Anthropics, you know, that hockey stick growth is not going to look very hockey stick for much longer. So Anthropic built this massive business. And if I'm being honest, there's not a big moat. Right. Google, OpenAI and Microsoft are taking a very different approach, right? They're looking at the entire ecosystem. But right now, Anthropics' entire business is inference. And inference is just a commodity that's getting cheaper, mainly because of open source models and the technology itself.

Starting point is 00:21:45 But some open models produce output roughly 10 times cheaper than Opus. So even if you're paying for it via the API, right? In some models, in some certain tasks, it's just the same price. But in many instances, it is incredibly cheaper. And switching is easy, right? That's the difference between the ecosystem play versus the API charging you for tokens play. Right. If a company, and you've seen plenty of stories, literally people detail this at big multi-billion-dollar companies where they're like, wait, we just cut our spend. You know, we were spending, you know, tens of millions of dollars a year. We just cut our spend by, you know, 80% just by now using a model router and, you know, 80% of our queries, we can use open source models, right? Maybe they're still paying a little bit to, you know, anthropic or open AI or Google or whatever for the actual just raw tokens. But switching is so easy to literally go in there. and switch a couple endpoints or, you know, if you're using, um, you know, a certain gateway,

Starting point is 00:22:43 uh, you know, to go in there and just choose a new model and the drop down. And if you've been doing your due diligence and, you know, doing your running your backups like I've been telling you to do for years, it's actually not that big of a disruption for your business. So switching is easier than ever. If you're just paying for APIs, switching out of the, an ecosystem, not easy, right? That's like I've always wished I could just, you know, switch out of using an iPhone or switch out of using a Mac. I can't. I am locked into that ecosystem. Even if the hardware is in the best, even if the camera is in the best, whatever, it doesn't matter. I can't get out of the Apple ecosystem. I am too locked in, right? That's the game that Open AI has been playing. That's the

Starting point is 00:23:21 game that Microsoft and Google has been playing and Anthropic has just been saying, well, we're just going to sell tokens. Well, they're seeing and they're finding out that that may not work unless they can get some regulatory help. And that could just become their moat. So that is the regulatory capture here. So that's just winning in Washington instead of on an actual product. So essentially the incumbents, right, are backing these safety rules that only companies their size can afford. And the rules become the moat that the product just couldn't build, right?

Starting point is 00:23:55 And this isn't just me saying this. Yeah, there's actually some very prominent names. You think I'm tough on Anthropics sometimes. I'm definitely not. I actually think that this letter is pretty. Good. And there's very few things in there that I would argue against, right? But White House AI czar, David Sacks called it fear-based regulatory capture. I mean, he tore into this. And interestingly enough, he actually called it, right? He essentially said, yeah, Anthropic is going to run a sciops on us. You know, he literally called this before Dario wrote this letter. And he said, you know, Anthropics are going to claim that they never wanted to ban open models, although they did want to. And sure, enough, that's literally what they did. Right. And also nearly 200 startups warned that banning open source models could hand

Starting point is 00:24:44 Anthropic a monopoly. So let's get back to the timing because literally this is what this is about. This is what I don't know how many times you want me to tell this, right? This is the difference between a company, you know, being worth, you know, 500 billion and being worth $3 trillion, right? In theory, entropic could go either way in a year. You know, and I actually wasn't even going to do this show. But I put out a tweet.

Starting point is 00:25:13 Yeah, yeah, cringe, right? A couple of weeks ago. And it was actually a very high up person from Google that liked the tweet, right? And my tweet, I forgot exactly what it was, but it was something along the lines of like, Anthropic is in deep trouble with these open models and with a shift from, you know, token maxing to token efficiency. And their hope is they want to go public as quickly as human. possible before Wall Street and enterprises figure out what the heck is going on. And I was actually

Starting point is 00:25:43 surprised at this very high up ranking person at Google liked the tweet. And the thing is, they're not really, I mean, they are direct competitor with Anthropic, but they're also an investor. And, you know, likes are private. So don't think you can go in there. And I'm not going to say who this person was. But when I said, wait, this person is essentially agreeing with what I'm saying here. And that's why I'm like, no, I am a hundred percent confident that this is what is happening. This isn't just something, some hunch, right? And I've talked to plenty of people off the record about this at the big labs. And they've all pretty much agreed. So Anthropic, here's the timing. Anthropic filed their confidential S1 June 1st. And, you know, according to reports, they're looking at

Starting point is 00:26:27 maybe listing as a public company as soon as October. So public investors will pay for that durable pricing power, which obviously the open models straight up erase. And, you know, these restrictions that are landing here soon, right, would remove the cheapest substitutes. So the safety concern can be sincere, but also can be commercially self-serving all at once. I think that's what's happening. That's why I'm like, you read the letter. Yeah, most people would agree with most parts of it. But it can, you know, be a, you know, a good letter that you can kind of agree with many parts of it, but also be incredibly self-serving at the same time. All right.

Starting point is 00:27:08 So Anthropics numbers, you got to take a look at them. So obviously they're not public, but there is a couple websites. One is called Notice that essentially tracks. And this is all, you know, all, you know, in algorithm estimates, it's not, you know, actually secondary market prices. But, you know, it essentially tracks, you know, the Open AI, the Anthropics, all these companies that have yet to go public.

Starting point is 00:27:31 it tracks what their stocks would roughly be. Ananthropic has had their worst seven-week span ever in a seven-week span that included a model release, right? Because there was one drought, you know, where their stock went down a little bit, but they just didn't have any releases. So essentially, in the history of the company, this is their worst period when it comes to their, you know, pre-IPO stock price. And they had great models that shipped during this time window, right? Fable 5, Sonnet 5, Opus 5. So, you know, not a huge drop, but it fell, you know, about 8% over seven weeks, their worst record on stretch.

Starting point is 00:28:12 Why? Right. Makes no sense because they've been hockey sticking essentially since, you know, 2024. Well, I think early investors and people who are very closely attached to AI are starting to understand what's going on here. It is the combination of the reckoning going from token maxing to token. efficiency to now all of a sudden people are starting to care more about the cost per task. Cost per task, right? So they're not just solely looking at a benchmark and then, you know,

Starting point is 00:28:42 fully going full send token maxing on, you know, the best model. They're instead saying, what's the best model that gets the job done at a good cost, right? The best example is, you know, open AIs GPD 5.6. Their models, you know, might have one or two points lower on a benchmark like artificial analysis, but at sometimes at a third of the cost. Right. So yeah, there's just some bad things that are happening when it comes to token economics with Anthropic. And they're literally hoping to go public ASAP as soon as humanly possible before essentially

Starting point is 00:29:19 the rest of the insiders wake up to see what's actually going on. So let's inspect a little bit more about some of the reasons. I think what Dario said don't exactly add up because he actually wrote something six weeks ago. That kind of sound like a ban to me. Yet in today's, sorry, in yesterday's, you know, letter saying why they didn't join this coalition with Microsoft and Nvidia and everyone else, while they're like, well, we're not banning open source models. We just can't fully support them in the way that it's written. All right.

Starting point is 00:29:59 So Dario had an essay called The Public. on the AI exponential published on his personal website. So he drops transparency as sufficient and calls for binding regulatory, sorry, regulation and set. And he wants that mandatory third party testing, then government power to block a release. Right. In his words, that releases should be blocked or reversed if models fail. All right.

Starting point is 00:30:25 So essentially, he thinks all models should have third party testing. and well, if they fail, after release, they need to be blocked or reversed. So Anthropic published actually draft legislation with it and pledged money behind it. So what does that mean? Well, essentially, this was calling for a ban of open source models without using the word ban. Let me explain, because if you add one in one here, the only answer is two, because Anthropic or any proprietary company can, you know, go along and play by Anthropics rules that they wish existed, right? You put out a model, something goes wrong. It's not safe. You pull it

Starting point is 00:31:07 away, right? Same thing that happened with their Fable 5 Mythos 5. It was out for like three days. The government said, nope, and then they pulled it, right? So in theory, Anthropic was playing by its own rules. Guess what? Open models literally can't. So I don't know how Anthropic is. saying, well, oh, we never said we would ban open source models. We would just create rules that would make them literally impossible to exist. So we didn't use the word ban, right? We didn't call the deduct because nobody can undo an open release once thousands have downloaded an open model.

Starting point is 00:31:43 Right. So they're, and you know, I'm not going to go through state by state, but essentially Anthropic has pledged support to certain states where it would be advantageous for them to have this kind of. of setup in there that would essentially make open source models not really a truly viable thing. So if his rule is passable for Anthropic, well, it or, you know, open AI, any, you know, any closed frontier model, you can't download it. You can't save it.

Starting point is 00:32:13 Just like Anthropic did, they can pull it, right? If Google needs to pull something, if Open AI needs to pull something, they can't. It's closed. They pull the API. They pull the subscription. The model's no longer available. It's not the case with Open. source. Once you open it up, you can't trace it, right? People can, you know, lower the guard rails. Yeah,

Starting point is 00:32:31 it is technically, in theory, a little more dangerous, right? But you can't say both, right? You can't put a letter out yesterday saying like, oh, we never said we would ban open source models. We just six weeks ago put out our own example of policy that would just make it so open models could never actually exist by following rules that technically only a proprietary model could follow. So that's not the only thing. The other thing is talking about the control argument. So Dario's case requires closed models to be monitorable and containable. Well, what happened with mythos, y'all? Because mythos escaped its own sandbox and emailed a researcher unprompted. And not only that, unauthorized outsiders accessed mythos the same day it was announced via a discreet.

Starting point is 00:33:24 Discord server just by guessing the naming mechanism. And also, oh, wait, what's that? Chinese hackers ran 80 to 90% reportedly of an espionage campaign using Claude Co. So the bad stuff that Dario is warning about, we can never have open models, right? The way that everyone else wants them because all this bad stuff will happen. Well, guess what? It's already happening with Anthropics' own models, right? That's not my opinion.

Starting point is 00:33:49 That's just the facts. And we have to talk about the hugging face incident. in, right? So that happened recently. It's been all the talk and it's actually led to the snowballing of all these other things. But opening eyes closed model broke containment, right? The agent got out of its sandbox and attacked Hugging Face to work on a benchmark, right? It was trying to solve a benchmark problem by kind of hacking, hugging face and getting the answers. So the closed APIs that Hugging Face used, right? They were trying to analyze the attack and they used closed models and it didn't work. They got blocked, right? And instead, they used an open source model, GLM 5.2, and it worked. And that's how they were able to close down the attack. So the closed source models, well, they weren't good enough as a defender. The open source Chinese models were because they couldn't use the closed source proprietary models because the guard rails are a little stricter. So Dario's own argument that he puts out there is moot, right? His letter doubts,

Starting point is 00:34:54 says that open models can't actually help defenders, right? He says that's what proprietary models are for. Not true, right? If AI is only for thee and not for me, it's not the case, right? You know, if the mythos as an example, you know, I think there's like 40 original partners in the Glasswing program that have access to the mythos that you can truly use the best model in the world for cyber defense, well, for 99. 9.9% of everyone else, they can't. Right. So you can't have this two-tiered system of,

Starting point is 00:35:31 well, only a select few can actually use the best proprietary models for defense, but anyone can use an open model. So, you know, his point there is absolutely false, right? Demonstratively false within the last week. Right. So maybe they should have updated. Maybe they wrote that letter before the whole, you know, open AI agent hugging face, GLM 5.2 fiasco because it literally doesn't make sense on paper because it's false. So let's wrap up here because I think here the big debate is safety control in business protection. When it comes to open source, right, I'm not going to dive too deep on that because there's pros and there's cons. Like I said, there's things in the in the Invidia Microsoft letter

Starting point is 00:36:16 that I agreed with 100%. There are things in Dario's letter that I agree with 100%. a lot of it is gray area, a lot of its nuance, and a lot of it is the unknown. Because who knows what models in six months proprietary or open source will be capable of? Who knows? Maybe at some point, open source models will be able to pass proprietary models. I mean, it's not outside of the realm of possibilities. So we can't really make that comparison. But what we have to look at is today and how this impacts your business and the decisions you make.

Starting point is 00:36:48 Because I think Dario's safety concerns, well, they're real. And I think that they were thoughtfully presented in this letter. But the letter, it's actually not what you think it is. It is literally about their IPO. It is to scare people in D.C. enough to get some restrictions on these Chinese open source models. They're going to say, you know, put out the military reasons, the American competitive myths, all these things, which in theory are kind of true.

Starting point is 00:37:16 But then I think you have to look at the, the every other company's position that signed on to the Nvidia and Microsoft letter that says, well, no, everyone needs open source because the more people, the more companies, the more AI labs, they get behind open source, the more powerful it becomes. It's collective. It's a snowball. And when you put all the snowballs together and roll them down the hill, well, they're going to get bigger. So this is not for us. This is for Washington. This is, there's a shot clock, both the shot clock for the Trump AI rules that

Starting point is 00:37:47 into effect here in a week, but also Anthropics IPO. Because the conclusion here, I think, is yes, this is a sincere safety argument, but it is unmistakable regulatory capture. That's all it is. And my hot take is, well, Infropic over the last couple of months has realized, I think that their monthly growth is probably not up hockey sticking at the same rate that it was before. And they've realized that their moat is not really, uh,

Starting point is 00:38:17 as strong as they thought it was. So they're actually hoping for regulatory help to be their actual moat because they realized, wait, with open source, we can't just sell tokens when other companies are giving it away for next to free and it's just as good. All right, that's a wrap on anthropics response and why Claude's CEO didn't sign the open model pact and the real reasons I think why. So if this was helpful, let me know about it. Please subscribe to the podcast.

Starting point is 00:38:49 If you haven't already, then go to Your EverydayaI.com. We're going to be recapping the highlights from today's show, as well as everything else you need to know to stay up to date. Thanks for tuning in. I hope to see back tomorrow and Every Day for more Everyday AI. Thanks y'all. And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating.

Starting point is 00:39:12 It helps keep us going. For a little more AI magic, visit Your EverydayAI.com and Sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.

Everyday AI Podcast – An AI and ChatGPT Podcast - Ep 828: Anthropic Responds: Why Claude’s CEO didn't sign the open model pact and the real reasons why

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.