US using Mythos for global cyber operations, Anthropic engineers helping on site: Report

The US National Security Agency is reporting using Anthropic's Claude Mythos to plan cyberattacks. The AI startup is believed to have sent at least half a dozen Forward Deployed Engineers (FDEs) to help the agency integrate Mythos. This comes despite the fact that Anthropic has previously had a fallout with the US Pentagon.

In Short

• US National Security Agency wants to use Mythos for cyberattacks
• Anthropic reportedly sends its engineers to help NSA use Mythos
• Claude Mythos is believed to be so powerful that it can hack any system

Anthropic’s Claude Mythos is believed to be one of, if not the most, advanced AI models out there when it comes to cybersecurity. Mythos has been able to find thousands of critical flaws in some of the most widely used software in the world. Anthropic claims that this model is so good that it cannot release it to the general public at least for now, due to how good Mythos can be when it comes to cyberattacks. Now, as per a new report, the US National Security Agency (NSA) could be using Mythos to plan cyberattacks across the world, and Anthropic is likely helping the agency use the model.

According to a report from the Financial Times, the US NSA may be using Mythos for cyberattacks. It is believed that Anthropic’s AI model could be useful in infiltrating networks of countries such as China or Iran.

Anthropic is helping the US to use Mythos for cyberattacks?

That is, the US may use the most advanced AI model to target its adversaries and exploit vulnerabilities in their systems, which may allow the US to get sensitive information or access to their networks. At the same time, Anthropic has reportedly sent around half a dozen Forward Deployed Engineers (FDEs) to the US NSA.

FDEs are responsible for helping clients deploy AI models into existing systems, and to customise the models based on their needs. These Anthropic FDEs are likely doing something similar – helping the deployment and customisation of Mythos for specific use cases in the National Security Agency, which may include the reported plans for cyberattacks.

One person, who is said to be close to Anthropic, said, "The best way to build a good defence is to build a good attack.” The person explained that if the US does not use Mythos for attacks, its adversaries may find a way instead. "If [Mythos] is not used to build attack agents, adversaries will find a way to do it," the person said.

Though it is unclear whether Anthropic’s engineers are directly involved in how Mythos is being used inside the NSA. There is also no official word on how the NSA may be using Mythos.

On X, some users were quick to react to this report. One person claimed that this was how governments planned to use AI – to exploit vulnerabilities. The person wrote, “Government's use of AI will not be to patch vulnerabilities, but to exploit them. But, yes, keep acting like Government will save us from AI”

Another user called Anthropic hypocritical for willing to work with the NSA after its fallout with the US Pentagon over unrestricted AI use. The user commented, “Makes Anthropic‘s objections to working with the Pentagon a bit hypocritical.”

Anthropic to expand access for Claude Mythos

Claude Mythos was initially available to only a limited group of US-based organisations when it was released in April as part of Project Glasswing. Ever since its release, Mythos has drawn concern from governments, financial institutions and IT companies because of its ability to detect and exploit software vulnerabilities. Countries around the world, including India have demanded fair access.

Earlier this week, Anthropic said that it was going to expand access to about 150 additional organisations in 15 more countries, including India. Though, the AI startup has not revealed which Indian companies will actually get Mythos access.

Do note that Anthropic has had a fallout with the US government in the past. Earlier this year, the Pentagon designated the AI startup as a supply chain risk after disagreements over unrestricted AI use. US President Donald Trump posted on Truth Social in February, “I am directing every federal agency in the United States government to immediately cease all use of Anthropic’s technology. We don’t need it, we don’t want it, and will not do business with them again!”

- Ends