www.it-connect.fr
The White House Wants to Keep Claude Mythos Under Control — Europe Still Locked Out
Faced with the power of Anthropic's Claude Mythos AI model, the US administration is opposed to the European Union being able to access it. Yet Anthropic said a few days ago that Claude Mythos would soon be available to a wider audience... It is hard to know what to expect.
The context around Claude Mythos
Announced last April, Anthropic's Mythos model sent everyone into a kind of frenzy, both companies and governments alike. The reason, as I have already mentioned several times, is its advanced cybersecurity capabilities, especially for spotting security flaws.
At present, Anthropic has deployed this model to a small group of companies and organizations as part of an initiative called Project Glasswing. If you follow the news, you know that it has already helped uncover thousands of security flaws.
A few days ago, on the occasion of the release of Claude Opus 4.8, Anthropic also specified that Claude Mythos would soon be available to the general public. But we may not have access to this AI in Europe.
The White House wants to keep control!
At the moment, Anthropic has not yet granted access to Claude Mythos Preview to the European Union, or to any other foreign government, with the exception of the UK AI Security Institute. Just one exception.
The explanation may lie, in part, within the walls of the White House. Indeed, following my previous article, some readers pointed out to me that the White House was opposed to Anthropic's plan to release Claude Mythos. After doing some research on the matter, I can confirm that there is indeed information available on this subject.
Officials in the US administration are opposed to Claude Mythos being made available in Europe, for national security reasons. An article published by CNBC mentions discussions between the European Union and the United States about Claude Mythos.
Thomas Regnier, spokesperson for the European Commission, told CNBC: "Cybersecurity is a shared priority and we have agreed to mutually recognize our respective standards in this area. At the same time, we are currently expanding technical discussions with Anthropic and other model developers who have notified their latest models to the AI Office."
That said, nothing is settled yet... CNBC says it obtained information from someone familiar with the confidential discussions between the EU and Anthropic. The AI giant reportedly told the European Commission that it would first need authorization from the US administration to access the evaluation version of Mythos. The problem is that the US administration seems to oppose this.
In reality, two camps are facing off:
- Europeans believe it is too risky to let the United States be the only ones to possess an AI like this. Claude Mythos can be seen as an extraordinary hacking tool: the NSA and the FBI can really have fun with it... Can't they?
- The United States believes it is too risky to open access to Claude Mythos to Europeans, even though we helped develop it with our data...
And let's not forget China: the United States very likely wants to preserve its advantage in the field of AI.
Claude Mythos is expected to become available to the general public soon, but who will actually be able to access it? That question remains unanswered for now, as there is too much uncertainty. Meanwhile, BNP Paribas will work with the French company Mistral AI to develop an AI model dedicated to cybersecurity: a Claude Mythos made in France?
