While the US government has banned Anthropic, the United States National Security Agency (NSA) is reportedly using its advanced artificial intelligence model to conduct cyberattacks.
This was revealed by the Financial Times, which reported that the AI startup founded by Dario Amodei is helping the NSA implement its powerful Mythos model for offensive cyber operations, embedding engineers within the agency despite the ongoing legal battle with the Pentagon.
In fact, already in mid-April the American outlet Axios had reported the NSA’s use of Mythos, despite senior Department of Defense officials – which oversees the NSA – insisting that the company represents a “supply chain risk.”
At the beginning of March, the Department of Defense officially designated Anthropic as a supply chain risk, stating that the company threatened national security. Anthropic, for its part, claims that the blacklisting is retaliation for its positions on AI safety. Such a label, which prevents the company from obtaining federal contracts, is usually reserved for organizations from hostile foreign countries, notes Afp. And the dispute has escalated to court, with the lawsuit still ongoing.
Thus, the “government’s cybersecurity needs seem to outweigh the dispute between the Pentagon and Anthropic,” commented Axios. Mythos is considered one of the most advanced models for finding cybersecurity vulnerabilities, so much so that its capabilities are the subject of discussions among various governments concerned about the possible applications of such a tool.
All the details.
WHAT IS HAPPENING BETWEEN THE COMPANY LED BY AMODEI AND THE NSA
According to the FT, citing two sources, the San Francisco-based company has placed about half a dozen employees inside the NSA as so-called “field-deployed” engineers to lead the use of the technology and customize the models for specific applications.
It is not yet clear whether Anthropic engineers are assisting the NSA in active operations. However, a person close to the situation stated that Mythos would be useful for infiltrating networks of nations like China or Iran, added the British financial daily.
“The best way to build a good defense is to build a good offense,” a person close to Anthropic told the FT, arguing that adversaries are probably developing their own AI-based offensive technologies. “If [Mythos] is not used to create attack agents, adversaries will find a way to do so.”
IMPLICATIONS RELATED TO MYTHOS
So far Anthropic has avoided making Mythos publicly available, arguing that its capabilities could also be used to develop cyberattack tools.
Earlier this week, the company announced the deployment of Mythos to 150 organizations in 15 countries, significantly expanding access to the system beyond the United States and the United Kingdom. Initially, the company had limited access to Mythos to about 40 organizations.
THE APPEAL LAUNCHED BY ANTHROPIC
Meanwhile, just yesterday Anthropic made an appeal to major entities, companies, and laboratories dedicated to developing artificial intelligence systems to consider slowing down the pace of development, suggesting that AI systems are evolving so rapidly that they may soon be able to improve themselves autonomously without human intervention, with potentially significant risks for society.
The possibility of slowing global AI development would be “probably a good thing,” the company stated on June 5 in a blog post. In the document signed by the head of the internal research institute, Marina Favaro, and co-founder Jack Clark, the company calls for the development of international mechanisms to slow or temporarily suspend the development of frontier AI, allowing safety research and institutions to keep pace with technological evolution.
In the post published by Anthropic, the idea is compared to the international regulation of nuclear weapons. Any suspension mechanism should also ensure that less scrupulous AI labs do not secretly continue to work to advance the technology during the pause.
