Anthropic, a private AI firm, has begun allowing users of its Mythos cybersecurity model to share threat intelligence, while simultaneously denying the European Union access to the same advanced technology. This refusal underscores the growing power of private capital in controlling critical digital infrastructure, even as state entities seek to leverage these tools for their own security.
Anthropic announced Mythos in early April as an AI model capable of rapidly identifying and exploiting vulnerabilities across the internet. Despite the urgent demand from both corporations and government bodies for tools to patch these bugs, Anthropic and OpenAI have withheld these powerful cybersecurity models from public release, citing fears of misuse by criminals or terrorists. This decision concentrates control over a vital defensive and offensive digital capability within a handful of private corporations.
The ascent of AI labs as major geopolitical actors was evident last week when President Donald Trump traveled to Beijing for a historic summit with Xi Jinping, where AI was a central topic. Trump's delegation included some of the United States’ most powerful AI executives, such as Elon Musk and Nvidia’s Jensen Huang, illustrating the direct integration of corporate technology leadership into state foreign policy and international power negotiations.
Capital's Grip on Critical Infrastructure
A continent away, the European Union has been engaged in an unsuccessful petition to Anthropic, seeking access to its Mythos model for cybersecurity purposes. This failure highlights the limits of state power when confronted with the proprietary control of advanced technology by private capital, even when the stated goal is collective security.
The Trump administration is reportedly considering testing or even licensing the most powerful AI models before their public release. These are moves the White House once labeled “dangerous” and “onerous,” indicating a shift in the state's approach to managing, rather than challenging, the power of AI developers. White House Chief of Staff Susie Wiles is spearheading Trump’s AI policy, publicly vowing to protect Americans from AI cyberattacks by ensuring “the best and safest tech is deployed rapidly to defeat any and all threats.”
This month alone, dozens of members of Congress have signed letters to the White House concerning AI regulation, signaling broad state interest in controlling the deployment of these technologies. However, a White House official stated that “any policy announcement will come directly from the President,” centralizing decision-making within the executive branch regarding this critical sector.
The State's Role in Managing Capital
The narrative of "safety" and "security" is used to justify state intervention, yet the underlying dynamic remains the protection of existing power structures and the management of technologies developed and controlled by private capital. The state's proposed solutions, such as licensing or testing, serve to formalize and integrate these powerful private entities into the state apparatus, rather than democratizing control or challenging the concentration of wealth and power.
Failed Liberal Solutions
The EU's unsuccessful petition for access, alongside the Trump administration's pivot from dismissing regulation as "dangerous" to now considering licensing, exemplifies the inadequacy of liberal reform efforts. These actions attempt to manage the symptoms of concentrated technological power without addressing the fundamental issue of private ownership and control over tools that are becoming essential for national and international security, thus extending the life of the current system without altering its foundations.