Recently, the conflict between American AI company Anthropic and the U.S. Department of Defense has escalated. Anthropic filed a lawsuit against the Pentagon, accusing it of misusing the supply chain risk label. However, legal disputes do not seem to have affected Anthropics cybersecurity cooperation with the Pentagon.
On June 4th, local time, the British Financial Times reported, citing two sources familiar with the matter, that Anthropic had sent employees to the National Security Agency (NSA) in the United States to assist the agency in deploying the Mythos model for launching cyberattacks.
The NSA is the largest intelligence agency in the United States, under the Department of Defense. Its main responsibilities include providing signal intelligence services and protecting the countrys information systems. According to sources familiar with the matter, about six employees from Anthropic were assigned to the NSA as frontline deployment engineers. Their role was to guide the use of AI technologies and customize models for specific application scenarios.
It is still unclear whether Anthropic engineers are assisting the NSA in carrying out actual operations. However, one person familiar with the matter claimed that the Mythos model could be used to carry out cyberattacks against China or Iran.
Sources familiar with the matter told the Financial Times, The best way to build effective defenses is to create effective attacks. The United States adversaries are likely developing AI-driven attack technologies. If we dont use Mythos to create offensive intelligent agents, those adversaries will find ways to do it anyway.
Regarding the related reports, Anthropic and the U.S. Department of Defense both refused to comment.
The headquarters of the National Security Agency, located in Fort Meade, Maryland, USA, AFP
Its important to return only the English translation. in the response. Do not include any Chinese text in your answer. Meanwhile, as Anthropic and the U.S. Department of Defense engage in AI technology cooperation, both parties are embroiled in a legal dispute. Previously, during negotiations for a contract worth approximately $200 million, there was intense debate over how AI tools should be used in wartime.
Anthropic demanded that the Pentagon ensure that its AI model, Claude, would not be used for large-scale surveillance within the United States, nor to power fully autonomous lethal weapons. The Pentagon believes that private companies have no right to formulate policies on behalf of the U.S. government.
After the negotiations collapsed, U.S. Defense Secretary Heggs announced that Anthropic was classified as a supply chain risk entity. On March 9 local time, Anthropic filed a lawsuit against the U.S. Department of Defense, accusing it of misusing the supply chain risk label to unfairly punish American companies.
After this dispute occurred, Anthropic launched the Claude Mythos model in April. This model was only available to certain US institutions that went through rigorous reviews. The company claimed that Mythos has excellent performance in the field of cybersecurity, and can identify network vulnerabilities on a scale that far exceeds human capabilities.
Anthropic claims that Mythos could reshape the landscape of cybersecurity. Currently, there are no plans to make it widely available. On June 2nd, local time, the company announced that it would expand the use of Mythos beyond the United States and the UK. The Mythos model will be distributed to 150 institutions in 15 countries.
The Financial Times pointed out that Anthropic has maintained close cooperation with the U.S. government, promoting its AI models within the U.S. government. For example, the U.S. Department of Defense is attempting to develop AI-driven tools for countering China.
On June 2, U.S. President Trump signed an executive order to establish a voluntary review framework. This framework allows the U.S. government to obtain access to AI models up to 30 days before their release, enabling prior security reviews. This order is not mandatory; companies can choose to participate voluntarily.
