Key Points
- US Treasury Secretary and White House chief of staff met Anthropic CEO on Friday
- Meeting signals potential thaw despite ongoing Pentagon court battle over supply chain risk
- Discussions centred on Anthropic’s Mythos model access for government cybersecurity
US Treasury Secretary Scott Bessent and White House Chief of Staff Susie Wiles met Anthropic CEO Dario Amodei on Friday, signalling a possible easing of tensions between the artificial intelligence company and the Trump administration. The talks came even as Anthropic remains in a court battle with the Pentagon, which had previously designated the company a “supply chain risk.”
The development is significant for India‘s fast-growing AI ecosystem, as many Indian companies and public institutions depend on US-led AI infrastructure and closely watch Washington’s regulatory approach. Decisions taken in the US often shape global access to frontier AI systems.
EVENT
Infosec Reimagined
Infosec Reimagined 2026 is the premier information security summit where top leaders—CISOs, CROs, CIOs, CTOs and risk executives—converge to redefine cyber resilience.
EVENT
CIO Prism
CIO Prism unites forward-thinking technology leaders to exchange transformative insights, shape digital strategies, and foster innovation, empowering enterprises to excel in an era of rapid technological change.
Anthropic, founded in 2021 by former OpenAI executives, develops large language models—AI systems trained on vast datasets that can generate human-like responses and perform advanced reasoning. Its products compete directly with OpenAI’s ChatGPT and Google’s Gemini.
Focus on Mythos access
A central issue in the meeting was how US government agencies outside the Pentagon could engage with Anthropic’s new Mythos Preview model. Mythos is designed with advanced cybersecurity capabilities and is said to identify vulnerabilities in software systems that human analysts may miss.
Because of its sensitive cyber potential, Anthropic has limited Mythos access to a select set of companies and organisations, including Apple, Google, Microsoft, Amazon Web Services (AWS), NVIDIA, Cisco, CrowdStrike, Palo Alto Networks, Broadcom, The Linux Foundation and JPMorgan Chase.
US Treasury and several other agencies are reportedly interested in gaining access.
Officials are concerned that tools like Mythos could eventually be used by malicious actors to target the US financial system. At the same time, such systems could help governments and corporations strengthen cyber defences before rivals obtain similar capabilities.
Pentagon dispute kept separate
Citiing a sources familiar with the discussion, Axios reported that both sides tried to separate the ongoing Pentagon legal dispute from broader discussions about cooperation with other government departments.
Advertisement
Wiles reportedly gave little detail on the Pentagon issue, noting only that the matter remains “in court”. However, she also made clear that the administration wants an open channel of communication with Anthropic.
The meeting also covered how Anthropic secures its code and how it decides when and how to release powerful new AI models.
Bessent joined the discussions because he viewed the matter as one involving a private company where government still has an important role to play.
An Anthropic spokesperson described the talks as productive, saying they focused on shared priorities such as cybersecurity, maintaining America’s lead in the AI race and AI safety. “We had a productive discussion on how Anthropic and the US government can work together on key shared priorities such as cybersecurity, America’s lead in the AI race, and AI safety.”
The White House also characterised the introductory meeting as constructive. “We discussed opportunities for collaboration, as well as shared approaches and protocols to address the challenges associated with scaling this technology.” It added that the talks also examined how to balance innovation with safety.
Why the Pentagon objected
Anthropic continues to challenge the Pentagon’s “supply chain risk” designation, which followed the company’s refusal to allow unrestricted military use of its software.
Amodei had insisted on guardrails preventing the technology from being used for mass surveillance or autonomous weapons—systems capable of selecting and engaging targets without direct human control.
Negotiations between the two sides eventually collapsed, and relations reportedly worsened during that period.
At the time of the Pentagon designation, President Trump said federal agencies should stop using Anthropic products. That order has not taken effect while the court challenge continues.
The company still faces internal scepticism within some parts of the administration. One official told Axios that Anthropic has used interest in Mythos to build support among receptive government figures and that the strategy appears to be working.
Another official said nearly every agency outside the Department of War wants access to Anthropic’s technology.
Your Questions, Answered
What is Anthropic and why is the US government meeting its CEO?
Anthropic is an AI company that builds large language models competing with OpenAI and Google. The US government met CEO Dario Amodei to discuss potential collaboration on cybersecurity despite an ongoing Pentagon court dispute.
What is Anthropic’s Mythos model?
Mythos Preview is Anthropic’s AI model with advanced cybersecurity capabilities. It can identify system vulnerabilities and has been released only to select organisations due to its powerful cyber applications.
Why did the Pentagon designate Anthropic a supply chain risk?
The Pentagon applied the designation after Anthropic refused to make its software available to the military for all lawful uses. CEO Amodei insisted on restrictions preventing use for mass surveillance or autonomous weapons development.
What happens next in the Anthropic-Pentagon dispute?
The court case continues with no immediate resolution expected. Meanwhile, other government agencies are exploring separate arrangements to access Anthropic’s technology, particularly the Mythos model for cybersecurity purposes.