In a significant development, Anthropic CEO Dario Amodei met with White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent on Friday, marking what the White House described as 'productive and constructive' discussions regarding access to Mythos, a cutting-edge AI model capable of identifying thousands of zero-day vulnerabilities. This meeting represents an initial step toward resolving the standoff that began when the Pentagon blacklisted Anthropic for maintaining safety restrictions, suggesting that any potential agreement would likely route Mythos access through civilian agencies, excluding the Defense Department.
During the meeting in the West Wing, Amodei discussed various opportunities for collaboration and shared approaches to addressing the challenges tied to scaling this advanced technology. Following the meeting, President Trump indicated he was unaware of the discussions, highlighting the covert nature of the negotiations.
The ongoing conflict has placed Anthropic in a precarious position as it remains blacklisted by the government, while officials scramble for access to its groundbreaking AI model. A resolution would likely see Mythos access limited to civilian agencies, avoiding direct involvement from the Pentagon, which initiated the blacklisting.
Background of the Conflict
The standoff originated in late February when Defense Secretary Pete Hegseth insisted that Anthropic grant the Pentagon unrestricted access to its AI models for various purposes, including military applications and domestic surveillance. Amodei's refusal stemmed from concerns about the reliability of AI for autonomous weapons and the inadequacy of current U.S. laws regarding mass surveillance. Consequently, Hegseth classified Anthropic as a national security supply-chain risk, effectively blacklisting it from government contracts.
In response, Anthropic filed two federal lawsuits against the Trump administration in early March, alleging illegal retaliation. While a federal judge initially blocked the blacklisting, an appeals court ultimately reversed that decision in April, leaving Anthropic excluded from Department of Defense contracts but available to collaborate with other governmental entities. Following this setback, the company engaged consultants from Trump's circle to help navigate toward a political resolution, culminating in Friday's meeting.
Interestingly, the backdrop to these negotiations is the recent announcement of Mythos on April 7, just days after losing the legal appeal. The model's capabilities have become too significant for the government to overlook.
Capabilities of Mythos
Mythos is a versatile AI model that has demonstrated an unprecedented ability to identify and exploit previously unknown zero-day vulnerabilities across major operating systems and web browsers. In testing, it successfully uncovered flaws that had evaded detection for decades and achieved a remarkable success rate of over 83% in developing working exploits on the first attempt. Notably, it became the first AI to complete a comprehensive corporate network attack simulation.
Due to its potential, Anthropic opted not to publicly release Mythos, instead creating Project Glasswing—a controlled access program allowing approximately 40 vetted organizations, including tech giants like Amazon Web Services, Apple, and Google, to utilize the model for identifying and mitigating vulnerabilities. The company has committed up to $100 million in usage credits for Mythos and is investing $4 million in open-source security initiatives, reflecting its commitment to safety and responsible AI usage.
Objectives of Each Side
The Treasury Department is keen on leveraging Mythos to enhance the security of its systems, with parts of the intelligence community already evaluating its capabilities. The White House Office of Management and Budget is working on frameworks to facilitate controlled access for federal agencies. Bessent's participation in the meeting underscores the importance of economic and financial security discussions surrounding Mythos.
For Anthropic, resolving the blacklisting is crucial. Although the company has achieved a valuation of $800 billion and is exploring an IPO, the supply-chain risk designation undermines its credibility and creates uncertainties for government-related partnerships. Amodei seeks a resolution that reinstates the company’s standing without compromising the safety protocols that sparked the original conflict.
Geopolitical Pressures
The international implications of this situation are pressing, as Anthropic is poised to provide Mythos to select British banks, further expanding its operations in London. The urgency is amplified by remarks from the Bank of England's Governor, who identified Mythos as a cybersecurity risk during a recent address. This development raises concerns that U.S. allies may access critical security tools before the U.S. government, adding a layer of urgency to the White House's negotiations.
Implications of the Meeting
The term 'introductory' used in the White House's summary of the meeting suggests that while discussions are opening, no formal agreements have been reached. The litigation remains active, and the appeals court ruling is still in effect. However, the fact that high-ranking officials engaged with the CEO of a blacklisted company signifies a notable shift in the administration’s stance, which seemed unlikely just weeks prior.
Amodei’s journey encapsulates a complex interplay of innovation, safety, and national security. As Anthropic navigates these challenges, the outcome of these discussions could set a significant precedent for the future of AI governance and its relationship with governmental oversight.