Anthropic has decided to pull Fable 5 from the market, its most advanced public model, just a couple of days after launch. In this way, Fable, which had quickly become the object of desire for AI experts around the world, disappears from our lives after a critical move for the company’s business in which, as has happened before, it seems that Donald Trump has had a great deal to do with it.
Before going any further, a bit of context: I had not personally tested Fable 5, so most of the references to its remarkable capabilities come both from Anthropic’s own documentation and from what some users who had gained access to it were saying on social media.
And it looked very promising.
More capability and greater efficiency in a model that arrived with the aura of being the mass-market version of Mythos, the AI that had made governments across half the world tremble just a couple of months earlier. So much so that Fable quickly rose to the top of the LLM Arena, the leading international benchmark for evaluating AI models.
Sam Altman does not like this
However, on Friday morning, the company led by Dario Amodei announced the withdrawal of both models, Fable and Mythos:
“The U.S. government, citing national security authorities, has issued an export control directive suspending access to Fable 5 and Mythos 5 for any foreign national, whether inside or outside the United States, including Anthropic’s foreign employees. As a result of this order, we must immediately deactivate Fable 5 and Mythos 5 for all of our customers in order to ensure compliance. Access to Anthropic’s other models will not be affected.”
So far, the Trump Administration has not been especially specific in detailing the reasons behind that suspension, which has only increased the sense of unfair treatment toward Anthropic, which has been in its crosshairs ever since it labeled the company a supply chain risk a few months ago and therefore excluded it from any type of government contract. All of this came after Anthropic refused to give in to some of Trump’s demands regarding the military and citizen-surveillance uses of its AI.
In fact, returning to Anthropic’s statement, that point becomes fairly clear:
“The letter (from the government) did not specify the reasons for its national security concerns. We understand that the government believes it discovered a method to bypass, or ‘unlock,’ Fable 5. We analyzed a demonstration of this technique and identified some minor vulnerabilities that were already known. These vulnerabilities appear relatively straightforward, and we have confirmed that other publicly available models also identify them without the need for jailbreak techniques.” (A clear reference to OpenAI.)
“To date, the government has only provided us with verbal information regarding a possible limited, non-universal vulnerability, which essentially consists of asking the model to read a specific source code and fix any software bugs. We understand that a potential vulnerability was shared with the government. We reviewed a report that we believe underpins the government directive and found that the level of capability shown in it is widely available in other models (including OpenAI’s GPT-5.5) and is used daily by system security professionals.”
Two references to Altman’s OpenAI in two paragraphs. Nothing further, Your Honor.
To make things even more tangled, some sources suggest that Amazon CEO Andy Jassy may have been the source of the security alerts that led the U.S. government to impose an export ban on Anthropic’s Claude Fable 5 and Mythos 5 models, forcing their global suspension.
According to reports from outlets such as The Wall Street Journal, Jassy reportedly told Treasury Secretary Scott Bessent that Amazon researchers managed to use the Fable 5 model to obtain information that could potentially be exploited in cyberattacks. Although an Amazon spokesperson declined to provide specific details, the company confirmed that it regularly advises the government on cyber risk matters and acknowledged that its own cloud platform, AWS, has been affected by the removal of these tools.
One more curious detail? Amazon itself is an investor in Anthropic.
Beyond the Fable case itself and however it is resolved from here, this situation leaves us with several lessons for the future, and they are not especially encouraging, either for the AI sector, for U.S. industry, or for the companies and users around the world who rely on these solutions.
The first, and most obvious, is that no matter how tempting it may be, do not put all your eggs in one basket. Especially when both the eggs and the basket may carry value as critical to your business as an advanced AI like Fable. And it is not only because something like this case can happen, with a government decision in the middle of it. One only has to look at the recent history of outages at OpenAI, Gemini, and others to understand that, as with other online services (hello, Cloudflare), the AI on which part of your operations depends may stop working without warning.
The second lesson, much more political, has to do precisely with the instability of the regulatory framework. We have always heard that companies, investors, governments… need a stable legal environment. A certain degree of “legal certainty,” allowing them to operate with the confidence that what they are doing will not suddenly become impossible the next day. In the Trump era, that legal certainty, that predictability about what will happen tomorrow, has been blown to pieces, as we have already seen in the case of his chaotic tariff policy. Today, the Trump administration is putting obstacles in Fable’s path, but… who can guarantee that the same thing will not happen to OpenAI in the future?
The third lesson has to do with geopolitics, with the game of cross-strategies that different governments around the world are playing, almost as if it were Risk, especially China and the United States.
Among the sources I regularly consult to understand how our sector is evolving is the technology section of the South China Morning Post, one of the flagship media outlets in the Asian giant. It speaks with some regularity, and with a certain sense of superiority, about the problems facing “Western” tech companies compared with the momentum Chinese companies have gained in recent years. That outlet was quick to comment on the Anthropic news, highlighting the contradictions of a U.S. government that champions economic liberalism while placing obstacles in the way of its own companies.
Many experts explain that, as of today, U.S. companies are still clearly winning the AI race, but no one can guarantee that this will remain true for much longer, especially given the momentum Chinese companies have given these technologies over the past year.
Finally, and for whatever it may be worth, I should say that my attitude toward AI regulation is, in general, favorable. I have followed the publications of the Future of Life Institute, which for many years has advocated for the conscious, regulated, and safeguards-based development of AI around the world, in order to avoid some of its greatest risks, such as those related to cybersecurity or the use of autonomous weapons in war. You probably remember the open letters that technology leaders of all kinds have signed about the dangers of these technologies when they are not developed within a safe framework. So I understand the precautions surrounding an AI like Fable 5, which was presented as being so powerful. But if that is the case, perhaps it should never have been launched in the first place.
And certainly, there should have been much greater clarity about the motivations behind a decision that, as we can see, alters the balance of the global AI sector.
Image: ChatGPT
Your email address will not be published. Required fields are marked *
Δ