US government met with AI manufacturers before Mythos preview rollout

US Vice President JD Vance and US Treasury Secretary Scott Bessent discussed AI safety with the heads of leading big tech companies. A corresponding phone call is said to have taken place last week, about a week before Anthropic's new AI model Mythos Preview was made available to selected companies. Mythos is said to be able to find security vulnerabilities in every important software and also develop suitable attacks.

According to the US news channel CNBC, Anthropic CEO Dario Amodei, xAI chief Elon Musk, Google chief Sundar Pichai, OpenAI chief Sam Altman, Microsoft chief Satya Nadella, CrowdStrike chief George Kurtz, and Palo Alto Networks chief Nikesh Arora were connected to the call. CNBC cites two sources who requested anonymity due to the confidential nature of the matter. An Anthropic spokesperson declined to comment on the matter but confirmed that a meeting with top government officials took place last Friday.

Prior to any external release, Anthropic briefed senior officials across the U.S. government on Mythos Preview’s full capabilities, including both its offensive and defensive cyber applications, he explained. “Bringing government into the loop early — on what the model can do, where the risks are, and how we’re managing them — was a priority from the start,” the spokesperson said.

According to the two anonymous sources, the call participants discussed the security situation with LLMs and their safe use, as well as how a possible reaction might look if the models develop in favor of attackers.

More banks are testing Mythos

As part of the newly presented "Project Glasswing" initiative, Anthropic according to its statements has made its new AI available to several selected companies. These reportedly include Amazon Web Services (AWS), Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks, among others. 40 other organizations responsible for software for critical infrastructure are also involved.

On the same day, Bessent and US Federal Reserve Chairman Jerome Powell convened an urgent meeting with the heads of the most important US banks to warn them about the dangers of Anthropic's new AI model Claude Mythos Preview. They appealed to the financial institutions to take the AI seriously. According to news channel Bloomberg, banks including Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley are among the companies with Mythos access. They have been asked by the US government to search for vulnerabilities with Mythos. Bloomberg learned this from people familiar with the matter.

In Germany, Mythos is being examined by the Federal Office for Information Security (BSI). They expect "upheavals in dealing with security vulnerabilities and in the vulnerability landscape as a whole," BSI President Claudia Plattner told dpa. If thought through consistently, there might be no unknown classic software vulnerabilities in the medium term, she believes. "This would lead to a shift in attack vectors and a paradigm shift with regard to the cyber threat situation."

(nen)