In the US, an extraordinary meeting of the president of the Federal Reserve, Jerome Powell and the minister of finance, Scott Bessent, with representatives of the country’s largest banks. In Britain, meetings of bodies such as the Bank of England and (NCSC) are off the schedule. And in India, inter-ministerial convocation, with clear mandates for preventive shielding of the financial sector. These and similar reactions are caused by the debate that has erupted around the possibilities of the “brainchild” of the technology company Anthropic, named .
Optimists and pessimists
The question that has been plaguing governments and government agencies since Anthropic announced in early April the risks stemming from potential use of Mythos remains the same. Since the program in question has the ability to find all kinds of vulnerabilities in all major operating systems and browsers, the main fear lies in its highly advanced hacking capabilities. “We need to bend over the mythos, what it means as a development for cybercrime,” noted Bank of England chief Andrew Bailey, speaking to . A Bloomberg report a few days later said that a small group of unauthorized users appeared to have gained access to the model, raising concerns about whether the company itself – which has launched the closed Project Glasswing alliance with leading technology companies – is able to guarantee any security.
The uncertainty surrounding the issue raises reasonable questions. On one side are the alarmists. As the Minister of Finance of Canada, François-Philippe Champaign, who, in the context of the meeting of the International Monetary Fund in Washington, said: “The issue is quite serious because, unlike known geopolitical risk points, such as the Strait of Hormuz, where the parameters are clear, in this case we are dealing with the unknown.”
Brussels appears troubled and active at the same time, with Commission officials having met at least three times with Anthropic agents, trying to gain access to the program, without having managed to secure an agreement. China and Russia have not publicly positioned themselves, but analysts estimate that the mere fact that they do not have a seat at the table is creating great anxiety in Beijing and Moscow. “For China, it’s the second wake-up call after” Matt Sheehan, associate of the Carnegie Endowment for International Peace think tank, points out in the New York Times. And he adds that the prospect of the US gaining a lead in the field of AI widens the technological gap in Washington’s favor too much.
On the contrary, in the UK, the only country after the US with access to Myhtos, they see positive prospects, beyond the individual risks. “This is an upgrade to the digital capabilities of AI technologies,” said Deputy Minister of Artificial Intelligence and Digital Security, Kanishka Narayan. It was preceded by a report by the AI Security Institute, where it was highlighted that although Mythos is a very powerful digital tool, its dangerousness is directly proportional to the vulnerability of the defending security systems. “We cannot say with certainty that Mythos would have the same success against well-defended systems,” it specifically states.
Changes and views
There are, of course, those impatient with the case. Like US Secretary of War Pete Hegseth, who has repeatedly called on Anthropic and its CEO Dario Amody to lift any restrictions on use by the Armed Forces. Hegseth’s admonition comes two months after he called Anthropic a “supply chain risk” and ordered all federal agencies to stop using the Claude model, after the company demanded that its AI models not be used for mass domestic surveillance or autonomous weapons systems.
The company responded with a lawsuit against him and the Ministry of War. But current developments dictate a change of course. “The new Mythos model keeps Anthropic on the doorstep of the White House,” technology editor Miranda Nazaro writes for the website, adding that Amodei’s meeting with high-ranking government officials seems to have gone very well. The change of attitude was emphatically reflected in what US President Donald Trump himself said: “We had very good discussions. They belong to the Left politically but we cooperate very well.”
It is this power of technology companies and their subsequent power to dictate policy changes that has some talking about the need for the US government to talk to multiple software vendors. “Variety of options is key,” Matt Mittelstadt, a researcher at the AI Policy and Strategy Institute, told The Hill.
Still others raise questions about the viability of an exclusively algorithmic economy. The American magazine Wired in an article months ago, amid the AI frenzy, noted, referring to MIT research: “The study from MIT suggests that larger and more demanding AI models may soon offer reduced returns compared to smaller models. The margins for growth in the next five to ten years will start to narrow.” Similarly, the ever-cautious most recently cautioned that in the recent past similar proclamations of innovation have been accompanied by the appropriate “salt”, intended to drive up the stock prices of tech companies.
