Fable 5 Shut Down by US Government

0:00In this emergency episode, we are discussing the U.S. government shutting down Anthropics Fable 5.

0:06NLW:The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.

0:18NLW:All right, friends, quick announcements before we dive in. First of all, thank you to today's sponsors, Robots and Pencils, Scrunch, Blitzy, and Out Systems. To get an ad-free version of the show, go to patreon.com slash aidailybrief. or you can subscribe on Apple Podcasts. If you want to learn more about sponsoring the show, send us a note at sponsors at aidailybrief.ai. And while you're there, check out the new site. If there is any specific part of a specific episode that you want to share with someone, whether it's a number or some stat or quote, there's a good chance that it is now there cut up and shareable for you.

0:48NLW:So go check it out, aidailybrief.ai. Hello, friends. Welcome back to the AI Daily Brief. For the first time in the three-year history of this show, news has broken on a Friday afternoon that is too significant to wait until Monday to explore. Last night, just before 9 Eastern time in the U.S., Anthropic tweeted, the U.S. 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 our customers to ensure compliance.

1:29Access to all other cloud models is not affected. We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible. After this absolutely stunning news, journalists and internet sleuths flew into a tizzy to try to figure out what the heck had actually just happened. The Wall Street Journal added some color, reporting that Commerce Secretary Howard Lutnick had sent a letter to Anthropic CEO Dario Amadei announcing that the new models Fable 5 and Mythos 5 were now subject to export restrictions, meaning usage by customers outside the U.S.

2:00as well as foreign nationals within the U.S. would be prohibited. So where did this seemingly capricious policy come from? It was apparently a report from another company about a jailbreak it had discovered. Anthropic gave more details in their blog post writing, We received the directive from the government today at 5.21 p.m. The letter did not provide specific details of its national security concern. Our understanding is that the government believes that 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.

2:33These vulnerabilities all appear relatively simple, and we have found that other publicly available methods are able to discover them as well without requiring a bypass. And basically from there went on to say that they just don't buy the U.S.'s logic. They point out that in the weeks leading up to the release of Fable, they worked with the U.S. government and many others to red team Fable safeguards for a significant amount of time. They pointed out that, quote, no testers have yet been able to find a universal jailbreak, a jailbreak method that can very broadly bypass the model safeguards. Indeed, they write, we suspect that perfect jailbreak resistance is not currently possible for any model provider.

3:05Every safeguard used in the industry is vulnerable to non-universal jailbreaks, which can elicit some cyber information in specific circumstances. and it is likely that universal jailbreaks will eventually be found in the future. They said, Given that perfect jailbreak resistance does not appear to be possible today, Anthropic adopted a defense-in-depth strategy with Fable 5. We aimed to make jailbreaks either narrow or very expensive to produce, and to combine this with thorough monitoring to quickly detect and shut down any successful attacks. Harkening back to the controversy of this week, they continue, This is also why Anthropic has required 30-day retention of customer data with Fable, a policy change that carries real costs for us with customers, but that allows us to research and mitigate jailbreaks.

3:45Importantly, they conclude, we have not even received a disclosure of a concerning non-universal potential jailbreak that led to a harmful result. The potential jailbreaks that have been disclosed to us are either entirely benign responses or are minor findings that provide no mythospecific uplift. To date, they write, the government has only given us verbal evidence of a potential narrow non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws. Our understanding is that one potential jailbreak was shared with the government.

4:12We have reviewed a report that we believe is the basis of the government's directive and validated that the level of capability displayed there is widely available from other models, including OpenAI's GPT 5.5, and is used every day by the defenders who keep systems safe. Given that, then they write, we are complying with the government's legal directive and are removing access to Fable 5 and Mythos 5 for all users. However, we disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people. If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.

4:46Now, the Wall Street Journal later added, the jailbreak research in question was done by researchers at Amazon, who used a series of prompts to get Anthropics models to provide them with information about a handful of security vulnerabilities. Now, one note, as people will clarify, is that although the Wall Street Journal reported that the research was done by Amazon, the journal did not report that it was Amazon who shared the findings with the U.S. government. Nick on X writes,

5:20Now, Prinz on X put together a bunch of different posts to try to put together something of a timeline of what happened. They argued that, contra to Anthropik's argument, that every safeguard used in the industry is vulnerable to non-universal jailbreaks, and they stated that clearly when they released Fable 5, quote,

5:39was approved. Now, Prins added that per Axios, the government contacted Anthropic to ask to pause releasing the models but was unsuccessful. Or as they put it, Anthropic told the government to pound sand. Now, it's hard to wrap our heads around just how consequential this is. Risha Sharma was one of many to point out that a huge number of Anthropic's technical staff, including no less than Andre Karpathy, are not US citizens, but instead hear on things like EB1 visas, meaning that even internally they are not allowed to interact with these models now. So where we sit, at least at 7.32am Eastern Time on Saturday morning, is that Fable and Mythos are not available to anyone right now, and you gotta think that there is a flurry of behind-the-scenes activity trying to resolve this as fast as humanly possible.

6:22So what does the chattering class think? Dan Robustus on X teed up pretty much the entire conversation when he wrote, Am I mad at Anthropic or the U.S. government? Both? Probably both. Yeah, it's both. So let's talk first about the U.S. government side, or specifically the what-the-hell-are-you-doing U.S. government side. Many pointed out that the specific pretext for this banning is incredibly loose. AI entrepreneur Bindu Reddy wrote, This is really stupid. The U.S. banned Fable just because it responded with information that is already freely available on the internet. Every other model can easily be made to respond to some silly questions about common security vulnerabilities or how to make drugs or whatever.

7:00The cluelessness of the government is astounding. In the Wall Street Journal, the CEO of cybersecurity firm Lettuce Security, Katie Muzores, wrote, Who at the White House evaluated this and thought it was a threat? It's a complete overreaction because this is exactly the kind of prompting that defenders would do. AI policy expert Dean Ball wrote, I can't tell if this is lawfare against anthropic in particular or extreme national security hawkery. Regardless, it's simply cartoonish. Council on Foreign Relations senior fellow Chris McGuire wrote, If the Trump administration is so concerned about access to advanced AI models, why is it not enforcing the export controls currently on the books on advanced AI models, or the export controls that would require a license to buy large numbers of AI chips to make these models?

7:40Now, to be clear about Chris's position, he later tweeted, I actually think targeted export controls on model access are prudent, but across-the-board controls on all countries on a single model without any warning is highly questionable. Export controls are a critical tool and an extremely powerful one. Used correctly, they have the potential to massively extend the U.S. lead in AI. Used incorrectly, they will stifle AI development. The Department of Commerce's export control strategy has been completely incoherent and sabotaging. It is sending powerful AI chips to China, not enforcing controls that would prevent Chinese smuggling, creating massive loopholes that allow AI chips to be sent to China, and preventing U.S.

8:15AI companies from releasing their own models. This has to stop. We urgently need a smart export control strategy that applies robust export controls to deny our adversaries access to advanced technology while advantaging U.S. companies. Commerce and BIS are consistently doing the opposite. If BIS doesn't understand how to use its authorities or what the implications are of its actions, then it needs to find some new personnel who can actually execute a competent export control strategy. The current one is incoherent and self-defeating. Many pointed out that it was also hypocritical. Emerson Brooking from the Atlantic Council re-shared a post from the White House Office of Science and Technology Policy from just a couple of weeks ago when they bit back at the New York Times after the NYT reported that President Trump had signed an executive order asking tech companies to give the government oversight of new AI models before releasing them to the public.

9:00At the time, the White House account wrote, Lazy and inaccurate reporting on this policy. The EO creates a process for frontier labs to voluntarily share cutting-edge cyber models in order to secure critical infrastructure and strengthen the government's own cyber defenses. We are not conducting oversight of all new models. And here's the money quote, as that level of government overreach would have chilling effects on free speech and innovation. Indeed, the policy seems so baffling. For example, as Dean Ball again put it, an administration whose posture is that we should export advanced AI chips to China, which also wants to ban Britain and every other non-American on earth from using our best models.

9:36I have no words. Yes, to many, the policy is so baffling that it feels distinctly personal. Presaging the next point we'll get into, Josh Pigford wrote, Anthropic has not done themselves any favors with their hyperbole over the past 6 to 12 months, but I also guarantee this has zero to do with national security. Now, adding evidence to that is when the Department of War CIO Kirsten Davies tweeted, We fully support POTUS and the Secretary of War in prioritizing national security and the security of our warfighters, DIB partners, critical infrastructure, international partners, and allies. Some things are simply more important than revenue cycles, clickbait, and pre-IPO valuation.

10:12Now, I don't know who approved that tweet, and it could just be Davy's opinion, but that level of animosity specifically targeted to Anthropic makes it seem like this has pretty much nothing to do with Fable 5 and everything to do with the relationship between Anthropic and the government. Well-known tech journalist Ashley Vance writes, This strikes me as so petty and dumb on the government's part. They want Anthropic to do their bidding and are willing to hold the whole country back as a result. Georgetown Law's Peter Harrell is pissed. I find it ridiculous, he writes, and un-American for the government to tell me as an American, I cannot use an advanced AI model because of a vague and non-public alleged security threat.

10:48We should regulate AI, but based on transparent and impartial rules and not 5pm on a Friday diktats. Joey Politano writes, all of the worst impulses of the Trump presidency on full display. No plan or strategy, everything reactive, arbitrary, and maximally invasive. Anthropic is just repeatedly being singled out because they have insufficiently bent the knee. Putting it even more dramatically, Lassana next writes, Trump really wants to kill both OpenAI and Anthropic, nationalize their tech, and then become the emperor of mankind. Now, as much animosity as there is towards the US government, frankly, most of the industry's scorn right now is being levied on Anthropic itself.

11:25AI builder Sarah Hooker writes, you have to be humble even when pursuing excellence. I think the arrogance with which Anthropic has pursued the latest release has universally landed poorly. It is presumed everyone else should just be grateful to touch the technology even if it is intentionally hobbled, and no one else should be given permission to develop the technology because it is too dangerous. Jeremy Howard, who's no hyperbolic ex-poster hater, wrote, I disagree with this decision and I don't like it, but also, how did Anthropic not see this coming? It is the obvious response to, this is too dangerous for anyone except us to use, since that relies on a premise, we are uniquely good, that almost no one agrees with.

12:01Daria Anutmaz, who has spent the last few days not being able to use Fable 5 because it knows that he is a biomedical researcher, wrote,

12:21Now, summing up the sentiment behind this was a three-panel cartoon that went viral. In the first panel, a concerned-looking Dario Amade says, this is the most dangerous AI yet. It could kill us all. It will destroy all global infrastructure. This can't be allowed to fall into the wrong hands. In the middle panel, Donald Trump says, okay, it's banned. And in the third and final panel, an apoplectic Dario says, you can't do this. Investor, entrepreneur, and writer Will Manitas was one of about a bajillion people to point back to an old quote from a recent Anthropic blog post that said, the government should have the power to block or deter deployment of the model if it is determined in light of third-party assessment to present unacceptable risks.

12:56Will points out, Dario, 48 hours ago, U.S. government should be able to block model deployment. U.S. government, export controls models. Dario says, not like that. Now, some trying to point out that Anthropic did actually try to address this in their blog post, saying, as we have stated publicly, we believe the government should have the ability to block unsafe deployments as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles. Still, even with that caveat, a whole lot of people felt like this was an F-around, find-out kind of moment for Anthropic.

13:28Investor Nick Carter wrote, I can't believe Anthropic comparing their product to nuclear weapons 800 times backfired on them. I am shocked. Author Take Him writes, So livid right now. Anthropic overhyped mythos scared the living daylights out of the clueless global politicians like Treasury Secretary Besson and ECB President Christine Lagarde, and stoked a regulatory panic that may set back the entire AI industry. Expressing the same sentiment, but with a slightly more dispassionate voice, entrepreneur John Ennis wrote, My opinion is that Mythos is the current best model, but not actually some world-changing dangerous model, and that Anthropic did their usual song and dance about safety, largely because they didn't have enough compute to serve it at scale.

14:05So, then they launched Fable because they still have to think about the IPO, but they are still somewhat compute-limited, so they put all sorts of restrictions on it. Around the same time, because they are trying to get regulatory capture and not because things are actually dangerous, Dario did more scaremongering and published his honestly confusing white paper that offered no real solutions. So finally they succeeded, they managed to freak out the government, their cynical plan backfired, and now it's a giant pain in the butt. Or as Bantag simply put it, they named it Fable and then acted surprised when it came with a moral.

14:38NLW:One thing I keep seeing in enterprise AI, companies hedging across every cloud, every model, every framework, or paying a GSI for a pilot that never ends. The team's actually shipping. They've picked a lane and they move fast. That's one of the reasons I like today's sponsor, Robots and Pencils. They've gone all in on AWS. They're an advanced tier and AWS pattern partner, and they ship production AI co-workers in 45 days. That's led to them doing some of the more interesting work I've seen on AI co-workers. And by that, I'm not talking about chatbots. I'm talking about actual agentic systems that sit inside a business architecture and do real work.

15:11NLW:That kind of focus matters if you're an enterprise leader trying to get something real into production, or an AWS rep trying to move a customer from interested to deployed. Request an AI briefing at robotsandpencils.com. One conversation with robots and pencils and you'll know. Quick question. When was the last time you actually visited a website to research something? If you're like me, AI pretty much does that work for you now. That of course raises a new question for brands. If AI is doing the discovering, researching, and deciding, who or what is your website really for? That shift in user behavior, the rise of AI bots becoming your most important new visitors, is what my sponsor Scrunch is taking head on.

15:47NLW:Scrunch is the AI customer experience platform that helps marketing teams understand how AI agents experience their site, where they show up in AI Answers, where they don't, and what's preventing them from being retrieved, trusted, or recommended. And it's not just visibility. Scrunch shows you the content gaps, citation gaps, and technical blockers that matter, and helps you fix them so your brand is found and chosen in AI Answers. Now, for our listeners, Scrunch is providing a free website audit that uncovers how AI sees your site, where there's gaps, and how you're showing up in AI versus the competition.

16:16NLW:Run your site through it at scrunch.com slash AI daily. Weekends are for vibe coding. It has never been easier to bring a passion project to life, so go ahead and fire up your favorite vibe coding tool. But Monday is coming, and before you know it, you'll be staring down a maze of microservices, a legacy COBOL system from the 1970s, and an engineering roadmap that will exist well past your retirement party. That's why you need Blitzy, the first autonomous software development platform designed for enterprise-scale codebases. Deploy the beginning of every sprint and tackle your roadmap 500 % faster.

16:46NLW:Blitzy's agents ingest your entire codebase, plan the work, and deliver over 80 % autonomously. Validated, end-to-end tested, premium quality code at the speed of compute. Months of engineering compressed into days. Vibecode your passion projects on the weekend. Bring Blitzy to work on Monday. See why Fortune 500s trust Blitzy for the code that matters at blitzy.com. That's B-L-I-T-Z-Y dot com. This episode of the AI Daily Brief is brought to you by OutSystems, a leading agentic systems platform built for the enterprise. Organizations all over the world are building, orchestrating, and governing agentic systems on the OutSystems platform and with good reason.

17:23NLW:OutSystems' open and unified platform allows teams to architect, deliver, and scale governed agentic systems with agility. Teams of any size and technical depth can use OutSystems to build, deploy, and manage AI apps and agents quickly and cost-effectively without compromising reliability and security. With OutSystems, you can rapidly launch ideas from concept to completion. It's the leading agentic systems platform that is unified, agile, and enterprise-proven, allowing you to accelerate growth, reduce operational friction, and deliver real enterprise impact with AI. OutSystems, build your agentic future.

17:59Now, one thing that I will note is that to their credit, the safetyists are not dancing around excitedly. Eliezer Yudkowsky, who I disagree with pretty vehemently most of the time, wrote, I can't tell today whether this ends up good or bad. International treaties to stop all further AI escalation would be a definite good. Things short of that, complicated. This has some bad aspects like selectivity and likely overrule, and good aspects like pushing against the psychology of, but no government would ever dare tell AI companies to do anything, so give up, or raising doubts that impede venture funding for ever bigger models.

18:29So please stop tweeting about how I must be celebrating this. I'm not one of the kids who immediately goes into overacted victory paroxysms about any hit on a perceived enemy. I care about the effect on where things end up a year later, and that's a little harder to know the first day, you know? And in fact, trying to figure out where this leads, days, weeks, months, or years from now, was where a lot of the conversation resolved. Aaron Levy stating almost the obvious bluntly writes, This is a big turning point for AI regulation. The government is starting to deem some models too powerful for certain uses, which creates a precedent for a range of possible controls in the future.

19:01I'm in the camp that this is unnecessary and we should be primarily regulating the use of AI as opposed to the underlying models. But equally, there are plenty of people who actually prefer this outcome. Either way, it's unlikely that we're going back to a world where the government doesn't have far more meaningful involvement in the rate of AI progress. Andrew Friedman wrote, This is a moment we'll look back on as a major turning point in AI. For years, people in this community have warned that AI policy would get weird, that these systems would grow powerful enough to put them on a collision course with our institutions, our economy, our governments.

19:30I don't know what caused this moment. I don't know what it means for future models. I can't tell if this is targeted specific to Anthropic, but I can't think of a more overt act of government intervention in our capitalist society in my lifetime. AI policy just got weird. Coining a new and I think important term, Sterling Crispin wrote, the worst thing about this fable situation is that it just created precedence for capability thought crimes and drew a clear line in the sand going forward. Are the next round of models going to need DOW clearance before release? New open source models? This is not good for progress.

20:00Daniel Jeffries wrote, we're seeing a speedrun of a hideous future play out. Nobody can build a business on this quicksand and uncertainty. If we continue with this wild, gibbering, fear-mongering, and the fear-based gated access, and if we create the regulatory capturing policies this insane and idiotic and incoherent fear pushes the clueless towards, we will absolutely guarantee that the future of intelligence gets built outside America. Brian Zhao writes, if this sticks, this means Americans will need proof of citizenship to gain access to models on the level of mythos. That means potential ID verification not just on Claude, but everywhere Fable is served downstream.

20:30Cursor, Devon, OpenRouter, etc. A law firm that uses Harvey serving Fable 5 will get impacted just the same. Brian also writes, how is Anthropics supposed to serve Fable through API billing? They will somehow have to figure out a way to verify citizenship of the end user. API access will need to be drastically changed before access, even to American companies and citizens can be restored. And of course, researchers at Frontier labs themselves will no longer be able to use their own models. Ryan points out also that Open AI and Google DeepMind no longer have incentives to ship anything Mythos caliber until this is resolved.

21:00If they release it, any company that can jailbreak the model can get export controls imposed on the model, and then they now have to deal with the same headaches. Plus, any non-U.S. partners with Mythos Access through Project Glasswing get cut off. Now that the U.S. has exercised its kill switch once, expect other countries to operate with the assumption that Frontier Access can and will be revoked unilaterally. Honestly, one of the moments that I'm reminded of in AI history was when Sam Altman was first removed and then reinstated as OpenAI's CEO. That was the beginning of the end of their relationship with Microsoft.