Anthropic's IPO filing on Monday set the stage for the Claude creator to pursue possibly the largest initial stock offering in history. According to CNBC, investors are measuring the 5-year-old company's potential market cap in the trillions, putting it in a category with the world's most valuable companies, even as the stock market is showing signs of jitters on the AI trade.
Anthropic devoted 94 pages of its 400-page securities registration statement to the "risk factors" section. "If misused, replicated, weaponized, or operated with a high degree of autonomy," Anthropic acknowledged, "it could pose catastrophic or existential risks to humanity."
On September 29, President Trump convened the CEOs of the leading U.S. AI companies and announced a voluntary “Joint Commitment on Frontier Responsibilities.” The agreement calls on frontier-AI developers to establish robust internal controls, use independent external auditors, and create board-level committees to review safety assessments. But since it is not legally enforceable, it is neither "morally binding" nor a “constitution” for AI safety.
The AI giants being asked to police AI admit that increasingly autonomous systems can behave in ways their creators did not anticipate. Yet, they are promoting the unpredictable products within the U.S. military-industrial complex. The net effect? A historical acceleration of existential risks.
Weaponizing U.S. technological primacy
Trump has repeatedly framed AI as a competition between the United States and China, saying there will be “a winner and a loser.” He has rejected cooperation with Beijing because he believes the U.S. is already leading.
But the administration is not just pursuing laissez-faire economics. It is constructing an AI-national-security-industrial complex.
In January, war secretary Pete Hegseth issued the Pentagon's milestone AI memorandum, which laid out a plan with a drastic focus on acceleration, not deceleration. “Speed wins,” Hegseth wrote.
In spring, according to CNN, a U.S. Special Operations Command analyst used a large language model that almost misled the U.S. into instigating a war with China.
And yet, the June 2026 national-security memorandum directed the U.S. national-security establishment to obtain the most advanced AI systems, rapidly integrate commercial and open-source models, build high-security computing infrastructure and establish a strategic reserve of AI expertise.The objective is a “decisive and enduring AI advantage” over adversaries.
This fosters an odd hybrid of market freedom, enormous public infrastructure, military integration and geopolitical protectionism, without legally enforceable safety regulation.
AI ecosystem is already transnational
In view of the economic rationale, frontier AI exhibits enormous potential first-mover advantages. So, Washington is unwilling to impose rules that might slow U.S. companies while China continues developing its own ecosystem - even at the risk of accelerating catastrophic risks.
The problem is that AI is already transnational.American companies depend upon global supply chains; advanced chips involve Taiwan, South Korea, Japan and Europe. Models circulate globally. Open-source weights cross borders instantly. Researchers move internationally. Cyber vulnerabilities are inherently transnational. And AI systems increasingly interact with infrastructure belonging to other countries.
The international community cannot accept an exclusively U.S.-controlled AI order.
China is simultaneously pursuing rapid AI deployment and a substantially more state-directed safety regime, including mandatory safety assessments and emerging standards for autonomous agents. Its 2025 Global AI Governance Action Plan explicitly calls for international cooperation, safety, controllability and inclusive participation.
Europe is moving in a different direction. The EU AI Act imposes obligations on general-purpose models and substantially stronger requirements on models classified as presenting systemic risk.
Trying to make the U.S. the sole gatekeeper risks converting an already competitive technological system into a fragmented geopolitical system. Instead, what is needed is international multilateral cooperation.
Atomic lessons to AI regulation
Following the success of the Manhattan Project, a fierce debate erupted over the military and civilian architects on who should control the newly unlocked power of the atom. General Leslie Groves, who directed the Manhattan Project, championed military retention, while J. Robert Oppenheimer, director of the Manhattan Project's Los Alamos Laboratory, and a coalition of scientists fiercely advocated for civilian governance.
Groves's stated priority was national security. He and the War Department viewed atomic energy primarily as a supreme weapon of war. They believed that maintaining strict military secrecy and keeping custody of the weapons stockpile was essential to deter foreign adversaries, especially during the emerging Cold War.
By contrast, Oppenheimer insisted on civilian leadership. As scientists, Oppenheimer and his peers knew that the atom held significant dual-use potential. Atomic energy was not just a bomb. It had immense civilian potential for medical research, electric power, and peaceful scientific progress. Military control would choke off open scientific inquiry. Furthermore, they hoped to prevent the transformation of the United States into a permanent garrison state.
Finally, the scientists believed in international cooperation. A military-run agency would signal aggressive intent to the rest of the world, making international arms control and treaties structurally impossible.
Atomic and AI Regulation
The atomic debate shares striking structural commonalities with AI regulation, though the modern landscape features unique complications.
Like nuclear power, AI is dual-use technology. While it powers autonomous weapons, it also drives medical breakthroughs, consumer software, and economic growth. Restricting AI entirely to a national security framework risks strangling open innovation and commercial progress. But unlike the 1940s, the state does not own the technology; private corporations do.
Hence, the effort of the Trump administration and Defense Secretary Pete Hegseth to push for a military/national security overview. The two seek to filter AI governance through a hyper-accelerated national security lens, as evidenced by the establishment of the four-star Autonomous Warfare Command (AUTOWARCOM). The logic centers on three tenets.
First, the "wartime speed" race against China: The administration believes that traditional, civilian-led bureaucratic regulatory bodies are sluggish, risk-averse, and counterproductive. In their view, the U.S. is locked in an existential technological arms race with foreign adversaries, mainly China.
Second, hostility toward private sector "guardrails": When AI firms like Anthropic refused to strip out safety guardrails that prevent their models (such as Claude) from being integrated into fully autonomous offensive drones or surveillance tools, Hegseth designated them a "supply chain risk" and blacklisted them. The White House views civilian ethical constraints as a vulnerability that hampers national security.
Finally, bypassing domestic regulatory agendas: By framing AI primarily as an asset of national survival and weaponization, the administration shifts the center of gravity away from civilian federal agencies (like the FTC or DOJ) and places it into the hands of military and intelligence entities like the newly formed Technology Support Office (TSO). This allows the Pentagon to retain unrestricted, unhindered use of the computing architectures.
2026 is not 1946
In the aftermath of Hiroshima and Nagasaki, civilian leadership won the battle on the control the of the atom. The Atomic Energy Act of 1946 (McMahon Act) established the U.S. Atomic Energy Commission (AEC) under strict civilian authority, dealing a heavy blow to the Pentagon's initial plans.
There were three reasons for the civilian triumph. Oppenheimer and the newly formed Federation of Atomic Scientists engaged in a smart, coordinated public relations and congressional lobbying campaign, framing atomic energy as a turning point for humanity, not just a military asset. Second, they were supported by congressional skepticism of military monopoly. Finally, even President Harry S. Truman supported civilian control.
Unfortunately, AI regulation is unlikely to benefit from similar tailwinds. Today, the leading AI scientists operate in the corporate sector, not in the public realm. Second, decades of forever wars have dramatically transformed Capitol Hill, due to the increasing share of congressional members with military, intelligence and security backgrounds. To win and retain their seats, most rely on financial contributions by the military-industrial complex and related industries.
To make things even more challenging, the Trump administration, blinded by faded fantasies of U.S. unipolar primacy in an increasingly multipolar world, rejects the idea of civilian control.
And so, the scene is set for catastrophic risks — in the name of national security.
