Despite being involved in a legal dispute with the U.S. Department of Defense, artificial intelligence company Anthropic is still assisting the U.S. National Security Agency in deploying its high-performance Mythos large model for offensive network operations and dispatching technical engineers to the NSA to implement the technology. Two people familiar with the matter revealed that the San Francisco-based company has dispatched about six engineers to the NSA as front-line field technicians, responsible for guiding the implementation of the model and customizing and tuning the AI model according to the NSA’s specific operational needs.
It has not been confirmed whether Anthropic engineers are directly involved in the National Security Agency's current cyber attack operations, but a person familiar with the matter admitted that the Mythos model can be used to penetrate the Iranian country's network system.
An insider close to Anthropic argued: "The best way to build an excellent defense system is to master advanced attack technologies. There is a high probability that opponents are also developing their own AI-driven offensive network weapons. If we do not use Mythos to develop attack tools, the enemy will also find ways to implement similar technologies."
The collaboration comes as the Silicon Valley start-up is in a legal battle with the NSA, which is affiliated with the Department of Defense, over the use of AI technology in combat.
Anthropic previously refused to allow the U.S. to use its Claude series of large models to carry out large-scale surveillance of U.S. citizens and develop deadly autonomous drones. This move led to the Pentagon's unprecedented inclusion of the AI laboratory in the supply chain risk list. Anthropic filed a lawsuit over this; if it lost the lawsuit, it would be forced to terminate all commercial contracts with U.S. military cooperation agencies.
After the conflict broke out between the two parties, Anthropic launched a large model of Claude Mythos. This model can independently discover and exploit software vulnerabilities, causing governments, financial institutions and information technology companies around the world to generally worry about the risk of abuse.
Earlier this week, Anthropic announced that it would expand Mythos licensing to 150 institutions in 15 countries around the world, completely breaking the previous restricted opening rules that were only open to a small number of institutions in the United States and the United Kingdom. When the model was first launched in April this year, it was only open to a limited number of institutions in the United States.
The above layout coincides with Anthropic's preparations for its IPO, and the company's target valuation is expected to exceed US$1 trillion, highlighting the continued rise in the commercial and geostrategic value of artificial intelligence in national security.
After the launch of Mythos, competing product OpenAI simultaneously launched large models with similar offensive and defensive capabilities. Industry experts warn: Top AI can not only be used by hackers to invade computer systems, but can also become a powerful cyber warfare tool for countries to fight against hostile forces.
In response to related reports, Anthropic and the U.S. Department of Defense declined to comment.
The company has collaborated deeply with the U.S. federal government on large-scale model implementation, and the Trump administration is promoting the deployment of this AI in various federal security agencies.
