Ep 801: Fable 5 Drama Updates: The latest Between Trump vs. Anthropic and How it Impacts You

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.

The Anthropic Fable-5 drama is not just a fight between the Trump White House and Anthropic. If you missed it, Anthropic's new Fable-5 and Mythos-5 models were pulled globally after the US government government restricted access over national security concerns. Anthropic, though, says the government overreacted, while officials say the model created unacceptable cyber risks. But the key twist that no one is really addressing is that Anthropic helped write the government's very arguments. I mean, for months, Anthropic told the market that Mythos was too powerful to release, all while at the same time preparing for its own IPO in begging AI companies to slow down while also releasing multiple new models? That's why today's episode matters beyond just one model launch and all of the drama surrounding it. This is the moment that AI safety messaging becomes a business liability. And if your company builds or thrives on using Frontier AI, you now need to ask a harder question, not just which model was best, but which model can I still use tomorrow, or which model can I actually rely on? So we're not only going to be breaking down all of the Anthropic Fable-5 and Mythos-5 drama, but also giving a little bit of guidance on what your company should be doing when dealing with the now government-scrutinized frontier AI models. So let's get into it. Here's the big picture if you missed what's going on. So Anthropic launched Fable-5 and Mythos-5 as frontier cyber risk models. And days later, the US government essentially caused a global shutdown of these models over foreign access and national security concerns. So Anthropic said the government in response, though, overreacted to a much narrower jailbreak claim. So this all stemmed from, you know, a jailbreak claim that Amazon found, because Amazon reportedly warned officials in the government after their researchers tested Fable's cyber safeguards and flagged a couple of jailbreaks. Then Anthropic's own warnings about Mythos made the crackdown easier to justify. That's because after all of this, Anthropic CEO— yes, yesterday— is doing interviews bragging with a smile on his face about how early companies testing Mythos said you should need a gun license to use it. Yeah. So that's the big picture. We're going to be tackling it all, but stick with me for the next 35 minutes. And on today's show, you'll learn why Fable was pulled, who triggered it, and what changed daily. We're going to do a day-by-day breakdown because it's a lot. You're going to know how Dario Amodi's weapons framing became ammunition for the White House. You're going to know why AI safety messaging can become a business and regulatory liability. And you're going to know what enterprise leaders should change about AI vendor and model strategy. All right, let's get into it. Welcome to Everyday AI. My name's Jordan Wilson. If you're new here, we do this thing every day. It's a livestream, unedited, unscripted, uh, as well as a podcast and newsletter helping everyday business leaders like you and me not just make sense of all of these AI updates, 'cause my gosh, there's a ton, but I help you focus on what matters to grow your company and career. So it starts here, but make sure you go to our website at youreverydayai.com. Sign up for the free daily newsletter. We're gonna be recapping not just not the highlights from today's show, but everything else that's happening in the AI news world. You shouldn't be spending hours a day trying to keep up. I do that for you. You just get to be the smartest person in your company when it comes to AI. All right, let's get to it. Fable 5 drama. There's a ton. Uh, right? So, uh, first we're gonna zoom out here and provide a little bit of context before we get into kind of the day-by-day updates. And I might have a hot take or two Thursday. Haven't done a Hot Take Tuesday in a while, so might sneak in a Hot Take Thursday. But if you're like, wait, what the heck is Fable? I thought this thing was Mythos. What's happening? We're gonna get into that. But first, uh, Fable 5 was publicly available for about 3 days, and Fable 5 was essentially a version of Mythos 5. So it's part of the Mythos family, which is this highest class of models from Anthropic. So Fable 5 essentially had some added guardrails and safety checks that Anthropic said made it, uh, more safe to use or safer to use for the general public. So it is of the Mythos family and there was still, uh, companies that had access to Mythos 5. So that is the one without all of these added guardrails and safety features. But this was only available for a select few partners in their project Glasswing. So now you know what the heck Fable 5 is, Mythos 5, you know the names. Uh, let's zoom out a little bit on the timeline that matters, and this is important. So essentially, Anthropic in April started to tease and hype Mythos as this new model they were working on, and it was, uh, gonna be too powerful to release. So yeah, this is the first time an AI company, uh, aside from like going back to 20, like the early 2020s, this is the first time an AI company since we entered the ChatGPT phase has actually said, here's the model we're coming out with in a couple of months, but hey, FYI, it is so powerful. Okay. So essentially Anthropic spent about 6 to 8 weeks, uh, talking about Mythos and, oh my gosh, it's so good. It's too good. The world can't handle it. Right? Uh, then on May 28th, follow me on the timeline here, Anthropic releases Opus 4.8 and reminds everyone that Mythos is on the way. Days later, June 1st, Anthropic files S-1 paperwork to go public, and the Mythos hype is everywhere. 3 days later, yes, I'm not kidding here, right after releasing Opus 4.8, Anthropic calls for an industry-wide AI slowdown. You can't make this stuff up. All right. Then, yeah, right? Yeah. 5 days later on June 9th, they release Fable 5. Connecting the dots here yet? All right. And then on June 12th, there is the Fable 5 and Mythos takedown and the whole drama in the circus starts. All right. This was literally Friday afternoon at like 5 PM, uh, right? They get word from the government that they have to shut down essentially Fable 5 and Mythos 5 because they can't, uh, essentially allow access to foreign nationals. Indianthropics says, well, there's no way we can guarantee who gets access, who doesn't. So they shut it down for everyone worldwide. So the whole, uh, right? This whole thing, they spend months hyping this model. Uh, they release another model, they file to go public. They say, hey, industry, slow down. Then they release another model that gets taken off the shelf. And here's the icing on the cake. My goodness, I, I don't understand the PR and the comms teams from Anthropic, but I think this is all part of their plan. Because yesterday Anthropic CEO Dario Amadi went on a long Bloomberg interview and said, literally with a smile on his face, I kid you not, uh, go watch it, it's weird. Uh, so he said, and I quote, some of the early companies that we gave this this being Mythos 2, that we gave this to, said things like, this is a super weapon. You should have to own a gun license to use it. Please don't release this. So yeah, after all of this drama, Anthropic continues to add fuel to the fire, not exactly taking their foot off the pedal and trying to work with the Trump White House on this. All right. Uh, hey, FYI, livestream audience, it's good to see ya. If you have any questions, Adam saying good morning, uh, Angie joining from Montana, uh, Miko joining from Tokyo. If you guys have any questions, any hot takes of your own, uh, on Fable 5, get 'em in now. I would love to hear. Uh, but let me just give you some real talk here as we start to unwind and hopefully bring a little bit of clarity because that's what I ultimately want to do mixed in with a little bit of rambling, I guess. But I think this is all premeditated, right? Look at what SpaceX did. Zoom out a little bit. There was all this talk a couple of months ago, right? As SpaceX, Anthropic, and OpenAI start to go public, right? And all this, you know, talk around these AI companies finally going public. You know, SpaceX originally, they were like, oh, this company could be, you know, a trillion-dollar market cap company. And everyone's like, wow., right? Um, guess what? All right, I'm, I'm checking this, uh, live here, uh, because I know that there was a lot of movement yesterday. So SpaceX went, uh, public just about a week ago. Uh, initial talk that, oh my gosh, SpaceX could be a, a trillion-dollar market cap company. Guess what? Their market cap right now is $2.5 trillion. All right. So there's been obviously a lot of talk with Anthropic and OpenAI, OpenAI, uh, going public soon. All of these decisions, they aren't happenstance. Everything is premeditated, uh, because in the months coming up to, you know, going public, every move you make has hundreds of billions or trillions of dollars of implications. All right. So a lot of people are looking at this and feeling bad for Anthropic and saying, oh my gosh, poor Anthropic. They're losing so much revenue every single day, but they're doing the right thing. They pulled this very powerful model. Y'all, don't get fooled. This is pre-IPO marketing and stifling competitors because essentially Anthropic planted this seed of, oh my gosh, right? Recursive self-improvement and future models are too, too powerful. We can't release them to the public. All these bad things are gonna happen. Look at this, the, you know, this model found, you know, thousands of bugs across enterprise software that the whole world use. Banking systems are gonna go down, right? They spent months painting this doomsday scenario. And now we're at this point where, okay, they released their most powerful public model, Opus 4.8. Then they released another one right away, but they hyped this thing up and said, hey, you should need a gun license to own this. This is super dangerous. Right? And now the government, uh, the US government here is a little bit more involved than they were, partly because of the, uh, President Trump's new executive order, uh, on AI that essentially, it's voluntary, uh, but, you know, all the big labs essentially said they're gonna play, they're gonna submit these models, uh, give the White House 30 days to essentially approve them, look at the safeguards before releasing them. Now, all of a sudden, that process is going to be a lot more difficult. And it almost seems like to me, especially with Dario Mati's, uh, interview with Bloomberg, essentially just sticking his tongue out at the entire industry, Uh, saying, yeah, we released Opus 4.8, which at the time is the most powerful model. Then we followed it up with Mythos and Fable, intentionally, you know, planted all these seeds. So now it's gonna be very hard for our competitors to release any models. But look, it's okay because we have Opus 4.8 out there in the lead, and we just made it pretty much not impossible, but extremely hard for our competitors to launch anything because now, right? Everyone has their eyes on all of these capabilities. All right, so quick background. This didn't just start with Mythos and Fable-5. So if you go back to February, this is when the kind of Anthropic versus US government feud started. So essentially, Defense Secretary Pete Hegseth demanded that all lawful uses be allowed for Claude, right? So the military wanted to use Anthropic's models for anything that they deemed necessary. Anthropic resisted, uh, and because of two things, uh, they wanted bans on the government using their models for mass surveillance and autonomous weapons. And the dispute centered on corporate guardrails versus the Pentagon's operational control. So, uh, the government set a February deadline and forced Anthropic to choose safeguard or contracts. Then March 4th, right? Anthropic said, no, we're not going to do this. And Anthropic received a letter from the Department of War. And essentially this deemed Anthropic a national security supply chain risk, which had never been given to a domestic company. Right. And Reuters reported that that label barred contractors in the Department of War, Department of Defense, from using Claude for military work. All right. So that's some of the background that you need to know. Then in March, okay, Anthropic said the supply chain risk was legally unsound and they planned litigation to get it reversed. All right. And then Dario Amati, Anthropic CEO, said the Claude still supported intelligence planning and cyber operations. So they're like, okay, you know, we still want the US military to be able to use, our models, just not for these purposes. So, uh, Reuters called this an extraordinary rebuke against an American technology company. I agreed with that at the time. And the dispute made June's fable shutdown feel technical and political, right? So a lot of people are pointing to this and saying, well, this is, you know, the US government just getting their kind of payback against Anthropic. Okay. Now that you have the background, let's get to the day-by-day developments. So this all started on Friday. So yeah, this is less than a week old, but essentially on Friday, Howard Lutnick, the US Commerce Secretary, sent Anthropic an exports control directive. So this is only 3 days after Anthropic released Fable 5 and Mythos 5 to the public. Like I said, not everyone could access Mythos 5, but anyone could access Fable 5, the version of Mythos with guardrails. Uh, so directive barred any foreign national access to be able to use Fable 5 or Mythos 5. And then Anthropic essentially said, we can't guarantee or even know, you know, who's a foreign national, who's not. They said, we can't do this. So the only thing that we can do is shut it down worldwide because they couldn't, uh, comply with the government's demands because they said they were impractical. All right, so Anthropic said the restriction covered foreign nationals inside and outside of America, which made it hard to essentially comply with. And Anthropic said they reviewed one technical finding, which was only a known minor security vulnerability. So Anthropic argued that a universal jailbreak standard would halt Frontier launches industry-wide. So here is what Anthropic said in response in their blog post. They said, the US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all of our customers to ensure compliance. Access to all other Anthropic models will not be affected. We received this directive from the government today. This is Friday at 5:21 PM Eastern. The letter did not provide specific details of its national security concerns. Our understanding is that the government believes it has become aware of a method of bypassing or jailbreaking Fable-5. We reviewed a demonstration of this specific technique being used to identify a small number of previously known minor vulnerabilities. These vulnerabilities appear— all appear relatively simple, and we have found that other publicly available models are able to discover them as well without requiring a bypass. All right. So that is what Anthropic said. So the US government did not release, right, like a full statement like Anthropic did. They did not publicize, you know, kind of their quote unquote demands. But what I will read here is essentially this was from the 13th, I believe. This is from David Sachs. So his kind of recollection, because this has turned into a he said, he said, drama, right? So, uh, David Sachs, if you don't know, this was Trump's former AISR, and he still, uh, reportedly has a, uh, very influential, um, spot even though he is no longer in his official government role as AISR. Uh, but a lot of reports said that he had a lot of input, uh, on the government's new, uh, kind of AI, um, executive order. With this 30-day voluntary review process. So I'm going to read what David Sachs tweeted, right? Since I just read Anthropic's response, there is no official thing from the government. So think of this as a pseudo kind of official government stance or explaining the other side of the story now that you just heard Anthropic's. So David Sachs tweeted, I've had a number of conversations with folks inside and outside government about the current situation with Anthropic, and here is what I believe to be true. As we know, Anthropic publicly released its Mythos class models earlier this week under the commercial name Fable. Fable is Mythos with guardrails, but if those guardrails fail, then you've exposed Mythos and its advanced cyber capabilities to people who shouldn't have them. Keep in mind that Anthropic itself widely promoted the idea that Mythos was a cyber weapon and needed to be regulated as such. They asked the government for regulation of Mythos and championed the guardrails on Fable. If there is a vulnerability, big or small, it is Anthropic's, responsibility to patch. A highly credible trusted partner of both Anthropic and the US government who was testing Fable came forward with a jailbreak of those guardrails. The admin asked Dario to fix the jail— the jailbreak or de-deploy the model. Dario refused. In their blog post, Anthropic defended its decision by saying the jailbreak isn't serious. That is not what the trusted partner and the US government believe, nor is it the kind of minimizing language consistent with Anthropic's brand as the AI safety company. It's difficult to fathom how they could claim a jailbreak allowing operability of a cyber weapon could be defined as not serious. In the past, Anthropic has always said that safety must be top priority and taken super seriously. In this case, Anthropic prioritized the continued offering of the consumer model over safety. In reaction, the admin issued the exports control. The admin did this reluctantly. It's been very surprised that Anthropic hasn't wanted to cooperate with a reasonable safety request, i.e., fixing the jailbreak issue. Anthropic's reaction is very much at odds with their branding and ethos as a safe AI research company. The admin's hope now is that Anthropic remedies the safety issue, the export control is lifted, and Fable goes back into general release. The admin wants all of this to happen as soon as possible. It is frankly bewildered that Anthropic hasn't wanted to comply with safety requests that it previously said were its highest priority. Those trying to misdirect and tie this action to prior Department of War Anthropic issues are wrong. The admin values Anthropic's technical capabilities and feels that this issue, while serious, should be easily resolved. The ball is in Anthropic's court. Yeah, pretty, pretty interesting and candid takes there from David Sachs. And yeah, you have to— obviously, I think you have to take Anthropic's statement with a grain of salt. You have to take kind of David Sachs rebuttal with a grain of salt. But regardless, I mean, we're going to talk about what happened at the G7 summit on AI here. But, you know, essentially President Trump sat next to Demis from Google DeepMind and Sam Altman from OpenAI. And then across the table sat Dario Amati. So you could say that there's, you know, that the prior bad blood didn't do anything. You could say that, you know, both sides are amicably trying to work this out because there are implications beyond just Anthropic's business model, right? This does impact open source AI. This does obviously impact, uh, the US's ability to continue to lead in AI with China closing that gap quickly, uh, both on the, uh, closed— or sorry, both on the open model side, the open source side, uh, which China leads in the open source race, but also with China's, uh, frontier closed models as well. Because as we saw this week, if you pay attention to our newsletter, very powerful open model GLM-5-2 is actually now better, uh, in the top available model in the world for front-end coding. All right. So it's important that this is not just an Anthropic versus US government issue. This is a US versus, uh, China. This is geopolitical tensions, right? I've said this before. Obviously, we have the tie-in here with the US government, uh, wanting to use Mythos-level models for military operations. So this goes far beyond, uh, right? Just you, your company using the best model available. This is literally geopolitical global superpower, uh, tensions that are being worked out in real time in front of us. Okay, so let's get back to the timeline. On Friday, everything is pulled. You have these statements from Anthropic. You have a, uh, highly publicized rebuttal, uh, from, uh, the former AISR in the White House, David Sachs. All right, so now let's get to as the details start to unravel. So on Saturday, uh, we found out that the company that brought this up was Amazon. All right, so, uh, according to reports, Amazon CEO Andy Jassy reportedly warned the Treasury, uh, Treasury Secretary Scott Bessant Thursday, which ultimately led to the Friday pulling of the model. So Amazon researchers said Fable prompts exposed restricted cyber attack-related information during tests. Amazon said the governments often seek its counsel on security risks, and Axios reported that Amazon's warning triggered that Friday White House scramble. What's also important to understand here is that, well, Amazon technically had a lot to lose, right, with this. They were serving the model, right, via AWS. So presumably making money, right, on how most, uh, you know, big hyperscalers make money. They essentially upcharge on the API side and make money on companies, uh, you know, using these models, uh, on the backend or via API. But also they're a huge investor, one of the biggest investors in Anthropic. Uh, so, you know, a lot of people said, well, Amazon has their own models. This is, you know, big tech company versus big tech company. I'd say no. Uh, right. And I think that's ultimately why the US government took this seriously from one of the, uh, most respected companies and most respected leaders in Amazon, CEO Andy Jassy. All right. So then on Saturday, uh, the government administration officials claimed, right? Yeah, this is when it started to get weird and kind of personal. Uh, right. So we heard in reporting that the government officials claimed, you know, recalling the details on Friday, that they were trying to reach Dario, but he was unreachable. At a wellness retreat. And we've seen reports on the contrary saying, yeah, this was personal, this was false, that wasn't the case. Right. But we're getting more details as this unravels. And Anthropic called the retreat claim via reports absolutely false and offered senior leaders to address this issue with the US government on Friday immediately. And a company source said Amodei joined calls within 75 minutes. So just over an hour later that he made himself available. And former White House AISR David Sachs said Anthropic at the time refused to fix Fable, which we just read his statement. All right, then come Sunday, we're getting more details, more rumors. This thing's spiraling. So, uh, there was a Fortune article that tied the crackdown to, uh, actually China and their cyber fears because Semaphore reported that officials suspected a China-linked group was the one that accessed Anthropic's Mythos model, uh, via bypassing the guardrails. So first we said, okay, it's just Amazon, you know, being the nice company and alerting the government. But then we saw reporting and said, nope, it was a Chinese company that accessed Mythos when they weren't supposed to. All right. And Anthropic said that China access was not raised during the White House calls. And Anthropic also said it blocks product access from inside of China. All right. And then we had Anthropic's response. So on Sunday we said, in an Axios report, and we talked about this on our Monday AI News That Matters breakdown, uh, that essentially Anthropic was rushing their staff to Washington to talk with the White House and they were meeting. Uh, but essentially, uh, Anthropic said that the virtual technical meetings had occurred since Friday and that during these meetings, the administration officials claimed that Anthropic had not engaged seriously enough beforehand. So, you know, you had all these reports coming out saying that, yeah, Anthropic says that they're trying to, you know, make this situation better, but they're kind of, you know, not taking it seriously. And instead of, you know, making some of these changes, they're instead trying to fight them and say, hey, you can, you know, access, you know, this type of information on other models. So why should we budge on ours? All right. And then Monday. Yeah, kept unraveling. Uh, so an Axios report said that Anthropic struggled to speak the administration's political language and kind of likened this, uh, as a fallout between Anthropic's ability to properly communicate with the Trump White House. So one administration source said Anthropic took the wrong fork repeatedly. Uh, officials connected Fable frustrations to, uh, the February military use dispute. All right. Then also Reuters reported that Anthropic's technical staff met with the officials, but Reuters said that the National Cyber Director Sean Cairncross also joined those meetings, and officials fear the model could aid foreign military intelligence users, and cybersecurity leaders asked Lutnick and Cairncross from the US government to lift the restrictions quickly. That obviously didn't happen. All right, so then let's go to Tuesday. I'm gonna skip ahead. A lot of this is just details, details. All right, but here's where it started to get a little interesting. AI moves too fast to follow, but you're expected to keep up. Otherwise, your career or company might lag behind while AI-native competitors leap ahead. But you don't have 10 hours a day to understand it all. That's what I do for you. But after 700+ episodes of Everyday AI, the most common questions I get is, where do I start? That's why we created the Start Here series, an ongoing podcast series of more than a dozen episodes you can listen to in order. It covers the AI basics for beginners and sharpens the skills of AI champions pushing their companies forward. In the ongoing series, we explain complex trends in simple language that you can turn into action. There's 3 ways to jump in. Number 1, go scroll back to the first one in episode 691. Number 2, tap the link in your show notes at any time for the Start Here series, or you can just go to starthereseries.com, which also gives you free access to our inner circle community where you can connect with other business leaders doing the same. The Start Here series will slow down the pace of AI so you can get ahead. On Tuesday, Axios reported that 150 security leaders backed reversing the restrictions. So it seemed like outsiders, uh, were kind of lobbying behind Anthropic, trying to get the US government to, uh, you know, step back on this. But experts warned that companies may strip the useful Defender capabilities from other models. But still, by Tuesday, all of this remained offline. All right. So we had some new developments yesterday on Wednesday. And this is where a lot of people said, okay, maybe it seems like Fable 5 might not come back online for a while because of a new Wired report. So the Wired report said that officials want Anthropic to address the alleged Fable vulnerabilities first. And the NSA reportedly concluded that Fable guardrails could be disabled by jailbreakers. So essentially this Wired report said that the US government wanted, uh, Anthropic to guarantee that future versions could not be jailbroken. And Anthropic and security researchers said that's not really realistic. There's always going to be the possibility for jailbreaks with any public, publicly facing, uh, generally available large language model, right? So Anthropic said Monday that the government's jailbreak concern just remains substantially overblown, and security experts said that stopping all jailbreaks is not technically realistic. Then this one, we'll see how this plays out because this kind of, uh, came out yesterday, uh, but this was probably not a good look for Anthropic. I don't know how much they had to say on the timing of this interview release, but we had a Essentially a 90-minute sit-down interview with Dario Amati come out yesterday, right, where he's going through and reiterating about the Mythos before release, how early corporate testers called Mythos a super weapon and that those testers said that users should need a gun license and then begged to delay the model. So again, terrible look. For Anthropic to have their CEO on the timing. Again, I don't know if they agreed to the timing of this report, if it was, you know, when it was originally recorded, when it was agreed upon to be released, but an absolutely bad comms take from Anthropic in the middle of this to essentially come off, you know, and Dario Bodhi laughing. Yeah., right, with a smile on his face, almost in a bragging type of way saying, yeah, we created this thing that's a bioweapon and ha ha ha, you should need a gun license for it. All right. And then where do we stand today? All right. We had this meeting of G7 leaders overseas, uh, you know, with, uh, the, um, heads of Google's DeepMind, OpenAI's, uh, Sam Altman and then Dario Mati from Anthropic, uh, all meeting with world leaders. at the G7 summit talking about AI safety as one of the big focus points. Nothing, you know, this was not as much about the elephant in the room, the Anthropic versus the White House. Uh, President Donald Trump did state that negotiation with— negotiations with Anthropic are going fine. But obviously with that Wired report, it looks like Fable 5 may, unless Anthropic reverses course, kind of EthPro decides to play nice with the US government, reverses on its stance, right? But it seems like the only way for Anthropic to get its Fable 5, Mythos 5 back online sooner rather than later is to essentially reverse course, play nice with the government. But it seems almost like they're reportedly doubling down. And the reactions have been harsh and wide. All right, so these are You know, before I give you my take and we start to wrap up this episode, now that we went through the entire timeline, I want to go over a little bit of reaction. So first, Alex Stamos, he is Facebook's former security chief and Stanford cybersecurity leader. So he did sign that cyber expert letter demanding that Fable Access be restored that I talk about. So the letter said that defenders lost tools while adversaries kept alternatives. So what he said is he argued that similar cyber capabilities exist in rival public models, and Stamos' camp wanted transparent evaluations, not an emergency Friday takedown. All right. So Yann LeCun, one of the most famous names in AI, that's Meta's former AI chief scientist now, who has his own company, AMI. So he kind of mocked Dario Amati's Dario Amodei's warnings about the dangerous frontier models. And he said that Anthropic essentially scared regulators until regulators actually believed them. And that's, you know, kind of my personal take, right? And why I started with my kind of hot take there that Anthropic might have wanted this to happen. And I'll lay out my case with the timeline. But again, this isn't just me saying this. Jan LeCun, one of the most trusted and respected names in AI. I know he can be a polarizing interesting figure, kind of with his takes on scaling laws and large language models and all those things. But regardless, one of the most trusted and well-known AI researchers came out and said that, yeah, Dario is, you know, all his talk here, not a good look in that they almost tried to scare regulators until regulators actually believe them. So if you hype about the dangers, well, that can become the basis for future restrictions. Uh, so he talked about how Anthropic framed Mythos and Fables as uniquely powerful and dangerous, and Lecun argued that the government simply accepted Anthropic's own premise, right? So he said, well, Anthropic, you made the bed. You spent months talking about how dangerous this is. You release it publicly, so you gotta sleep in the bed you make. Uh, and his critique was not pro-government, it was anti-safety theater. So a lot of prominent AI voices have accused Anthropic, Dario, and others— of just safety theater. And I would agree with that.

All right.

Then you also have Jeremy Howard, another very respected name in AI. So he was formerly Kaggle's president and chief scientist and then founded Fast.ai. So he opposed the ban but called Anthropic's surprise unbelievable. So his take, he said that it is too dangerous for anyone except us, right, invites the restriction. So he said that there's an exceptional risk messaging that hands regulators control over your roadmap. Uh, then you have, uh, Dabby Ottenheimer, uh, who leads trust and digital ethics at Inrupt. So he questioned whether Mythos was truly as dangerous as advertised. I'd agree with that, right? If you go back to their original FYI, their original project Glasswing, blog post, which they've edited this a few times since the model has been released. But their big stat that they've led with was, was this CyberGym benchmark, right? It essentially measures a model's ability for cyber, right? Which is one of the big sticking points, cybersecurity, cyberattacks, all these things, right? If these models are so good at cyber, better than all humans, you know, they can take down banking systems, regulatory systems, all these things, and, you know, cause all this chaos.

Yeah.

Uh, right. Anthropic's own model, uh, uh, promotion talked about this CyberGym benchmark and how it was so far ahead of everyone else. All right. Couple weeks, couple weeks later, Microsoft released M-Dash. All right. And this is essentially a, uh, mixture of models approach and their score on CyberGym with off-the-shelf models. Crushed Project Glasswing's Mythos Promotions, right? So the main benchmark that Anthropic even used to justify how powerful their model was got absolutely smashed by Microsoft researchers in their M-DASH technology that just used other companies' off-the-shelf models. So, you know, it kind of— smart people that follow these things are like, Wait, is Mythos really that good? Right? It was, I think, a 6% increase. If you looked at Artificial Analysis Index, it was a 6% increase in terms of overall, uh, intelligence. So it wasn't like this, you know, 30% jump. It was a standard, right? Usually when a, a, a new, uh, uh, either OpenAI, Anthropic, um, or Google come out with a new model, if you look at the artificial intelligence, uh, Artificial Analysis Intelligence Index, right? Which is this conglomerate of all the different benchmarks puts, put together and it gives them a score. Usually the increase, the percentage increase is about 3 to 5%. So when Mythos, uh, Fable came out and it had that 6% increase, standard, right? So, uh, Davey, uh, Davey Einheimer said that this is a marketing trick. So he questioned whether Mythos was truly as dangerous as advertised., and his take was Anthropic used the security language to manufacture premium positioning. All right, so as I make my case for this, I want to again rewind to the timeline here. This is Anthropic's decision. This is Anthropic's timeline. All right. Regardless of if you're a huge Anthropic fan, if you're anti-Anthropic, if you love Google DeepMind, DeepSeek, whatever, I want you to think about this timeline again one more time. All right. So again, May 28th, Anthropic releases Opus 4.8, and then they remind everyone Mythos is on the way. All right. So they, uh, release Opus 4.8, May 28th, June 1st, couple days later, they say we're going public. 3 days later, they put out this blog post essentially saying, yeah, hey, everyone in the AI industry, we need to slow down, right? No more big models. 5 days later, they release Fable 5. Also, this is important because again, these rollouts have $100 billion or trillion-dollar market cap implications.

Right.

I can guarantee Whether you're talking SpaceX, which I think did a great job in the lead up to their IPO, right? Signing the compute deal with Anthropic, with the Cursor acquisition. I mean, SpaceX did this. They put on a masterclass in pre-IPO timing, right? The initial talk was, oh my gosh, $1 trillion. Here we are, you know, more than $2 trillion valuation for the company. so these things are scripted. Everything to a T. If you think that Anthropic was, uh, thrown off, that Fable 5 and Mythos 5 were pulled, you are the fool. They were not. My, my take, you know this when you spend months, which has never, these things have never been done. A company has never promoted an upcoming model and spent months going on press tours talking about how powerful it is. You know, oh my gosh, 50% of white collar jobs are gonna be be replaced, right? Everyone's saying that, you know, Anthropic is trying to turn themselves into a bad guy. Um, not me. This is everyone else speaking. But that timeline, this is the first time ever since the original ChatGPT launch that one of the three big companies, Google, OpenAI, and Anthropic, have ever released a more than a step change upgrade before competitors. So the first time, uh, when we talk about Anthropic releasing Opus 4.8, and then about 2 weeks later, uh, releasing Fable-5. This had never been done where one company had a sizable upgrade model. Normally Anthropic releases, then OpenAI releases, then Google releases, and it goes around in a circle. There's literally a meme that says, you know, Anthropic releases the most powerful model. 2 weeks later, OpenAI releases the most powerful model. 2 weeks later, Google releases the most powerful model, and it goes around. Anthropic, first time ever, had back-to-back releases without one of their competitors releasing something in between. And in between all that, oh, they filed to go public, and oh, they put out this industry call that we need to stop regulation. My argument is they absolutely knew that Fable 5 would end up like this. They had a preexisting— well, da— everyone, everyone.

Yeah.

Everyone knew that Anthropic and the US government and the Trump administration did not get along. You spend all this time hyping up this model as a bioweapon, its cyber capabilities, you should need a gun license. You know it's gonna get pulled. You know it's gonna get pulled. But they were fine with that because right before they had put out Opus 4.8, right? So here's my hot take. This is not temporary. Fable 5 access will be restored at some point, whether that's in a few days, a few weeks, I don't know. But I think ultimately Anthropic is playing 4D chess while the US government and all consumers are thinking in 3D checkers. All right. They are doing this for their IPO, period. Right? They're gonna play the card and people are gonna say, oh my gosh, you know, Anthropic, they're the good guys here. You know, they warned us, they put out this model, it's too powerful. So they did the good thing and they pulled it for the safety of all of us. Mm, no. All right. They knew this was happening. They know that ultimately whenever they re-release this, this is gonna be, uh, presumably weeks before they technically go public, right? We'll see if they follow SpaceX's timeline, uh, in terms of the time between their S-1 filing and actually going public. But presumably Fable 5, Mythos 5 is going to be re-released to the public. Anthropic will go public shortly thereafter. OpenAI and Google are going to face increased scrutiny, right? Getting their new models to market. I mean, who knows? We may see a GPT-5.6 today. Maybe it'll be another couple of weeks with Google's— Yeah. Gemini 3.5 Pro, but regardless, it, they intentionally made it, I think, harder on their competitors to get frontier AI out in the open. They created all of this drama. They built the roadblocks knowing with their back-to-back release cycle that they were fine, right? They're trying to get you to focus on, oh my gosh, you know, Anthropic was probably printing Monopoly money with Mythos 5 and Fable 5, and now it's off the shelves. Poor Anthropic.

No.

Again, the back-to-back release, which they've never done. They knew exactly what they were doing. And if you don't see through this, you are playing the fool, right? There's this meme that's been going around, you know, here for our livestream audience, you can see it. But, you know, essentially there's a little character of Anthropic talking to a computer. The computer is the US government. It says, regulate me, right? And then the US government says, you are being regulated. And then the Anthropic is OMG, I can't, right? I can't believe this. And they're The other one of someone, you know, riding a bike and, you know, sticking the, uh, the, the stick in between the wheels and saying, ow, you hurt me. Right? It's like, if you can't see through this, you are the fool. Right? Anyways, let's wrap this up. What should your company do? That's ultimately what this is about. I wanted to give you the backgrounds. I got in my personal opinion, my rambles, but I think the important takeaway here, Is you have to spend more time with the models that came pre-Fable 5, right? So spend more time with Opus 4.8, GPT-5.5, and Gemini 3.5, at least the Flash version, right? The Pro version could be coming soon, as those are more likely to be stable for the short, medium, and long term. Technically, I think so many companies spend too much time, and we talked about this on our 800th show yesterday, which, by the way, go give that one a listen. So I went a bit, a little bit more on this, but I think so many companies are leading, leaving huge ROI gains on the table because they're just model chasing, right? You should spend more time once you know and you've gone through with your use cases, you've figured out ROI on a certain workflow, right? Let's just say it's GPT-5.4 or the, you know, Opus 4.6. Okay. Spend more time rolling that out, getting that stable, rolling that across your organization. I think companies are maybe spending a a little bit too much time. Um, and just the capability gap is, I think, uh, really hindering a lot of companies' growth. So spend more time with the models that are stable. Number 2, you have to understand that enterprise AI access can disappear now through government action, not just vendor failure. All right? With now this, uh, you know, scenario that we have between Anthropic and the US government. So right now, your critical workflows you need fallback models, you need export checks, and you need continuity planning. And you have to separate the mission-critical workflows from the, you know, cutting-edge AI models. All right. So you can't essentially, if in 3 days, if the 3 days when Fable 5 was public, if you moved all of your mission-critical business processes to Fable 5 and, you know, kind of deleted your old AI game plan, you could think, oh, that's being AI native. You have to be quick to, to stay ahead. Sure. But you need to take the time to test those. And you should always have backups and redundancies in place because now at any point, if you put too much, um, of your company's day-to-day operational, uh, necessities in a model that could be pulled, you're only screwing yourself. So you need to spend all this time that you win from, you know, increased productivity,, on creating and building modularly. All right. So that's a wrap. One question here. Adam says, do you think that's why the June 22nd date, the way users had a roundabout heads up this would go away? So yeah, uh, essentially when Anthropic initially released, uh, Fable 5, they said you're gonna get access to this to a short period of time, uh, if you are on a subscription only through June 22nd. I honestly think that was just to, uh, get people to use the model as much as possible, hopefully become reliant on it, uh, and then start paying the API pricing. All right. So, but that's a wrap. Now, you know, at least the timeline of the, uh, the drama between Anthropic and the US government. I don't think this is gonna end today or tomorrow, but it will end soon. So I want you, the smart business leader, if you are in charge of AI strategy at your company, company. You cannot take, uh, right, a, a blog post at face value. You have to understand the bigger play here, not just with the geopolitical implications of something like this, uh, what this means for open source and what that can mean for your company. But you have to also look at what these AI companies are doing pre-IPO. Uh, this is not a mistake. This is not something where, oh my gosh, they're caught off guard. Everything is intentional. So you, in the same way that these companies are intentionally mapping things out, you need to do the same thing when it comes to using frontier AI models. If you play too much on the edge, you're gonna get cut. So you have to have a backup plan and redundancies in place. And that's what we help you do here at Everyday AI. I hope this one was helpful. If so, please subscribe to the show if you haven't already. Then make sure to go to youreverydayai.com. We're gonna be recapping this, uh, right? So you're gonna get this shorter. too long didn't watch version in our newsletter. So thank you for tuning in. Hope to see you 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. It helps keep us going. For a little more AI magic, visit youreverydayai.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. Bye!