Even the creator of a new AI is scared to release it to people. Because this tool is so strong that it is scaring them. Anthropic, the makers of the popular AI assistant Claude, has created a powerful version called Claude Mythos. And it has left governments, banks, and tech companies deeply worried.

What is Claude Mythos?

For the very first time, Anthropic has recently announced a preview of Mythos. It is a novel edition of its AI platform, Claude. In the recent preview, Mythos achieved a good score in all areas. The company claimed that Claude Mythos is brilliantly amazing at coding. Particularly in security-related assignments.

This is the shocking part. During the testing, Mythos Preview detected bugs in all the major operating systems and web browsers. Some of the bugs, including those considered decades old, were not found by human security experts. Imagine it as a really intelligent detective. However, rather than preventing crimes, it discovers back doors in computer systems, the back doors that can be easily exploited by hackers.

Why Did it Stay a Secret for So Long?

What Anthropic did was highly unusual. According to Anthropic, its latest software cannot yet be presented to the public.  One of the main reasons is that it works very efficiently in identifying any serious problems in the system. As such, there is a risk that this tool will be used by hackers and other dangerous agencies to harm other individuals.

Instead of presenting it to anyone, Anthropic plans to give access to the preview to 40 companies. Those companies will use this model to scan and protect their own code. A few organizations, including Apple, Google, Microsoft, Amazon, and JPMorgan. Mythos is used by these organizations to detect any problems and fix them before the hacker gets a hold of them.

Why are Banks and Governments are concerned?

This issue has surpassed the field of technology. A meeting was held between the CEOs of the largest U.S. banks and the Federal Reserve Chair Jerome Powell and Treasury Secretary Scott Bessent. The main motive was to discuss the possible cyber threats of the Mythos model by Anthropic. 

The bank CEOs were already in Washington for a financial services forum meeting. The Mythos emergency meeting was called on top of that. The room was filled with CEOs of Goldman Sachs, Wells Fargo, Morgan Stanley, and others – though JPMorgan’s Jamie Dimon could not attend. 

Even some theorists explain that AI can assist hackers in crashing the financial systems. It can even put a pause on hospitals and manufacturing plants. It can help one country to destroy the infrastructure and everything in another country.’

The Anthropic Claude code tool, which is already publicly available, was already used by a Chinese state-sponsored group to attack about 30 organizations in a well-coordinated attack before its discovery by Anthropic.