A recent high-stakes confrontation between the Trump administration and San Francisco-based AI company Anthropic has brought the contentious debate over AI governance and ethics to a head. The Defense Secretary, Pete Hegseth, invoked a national security law to blacklist Anthropic from Pentagon contracts, a move triggered by the company's refusal to permit its AI technology to be used for mass surveillance of U.S. citizens or for autonomous armed drones capable of selecting and eliminating targets without human intervention. This unprecedented sequence of events could cost Anthropic a contract worth up to $200 million and potentially bar it from working with other defense contractors, following President Trump’s directive for federal agencies to cease all use of Anthropic technology. Anthropic has indicated it will challenge the Pentagon's decision in court.

MIT physicist Max Tegmark, founder of the Future of Life Institute and a long-time advocate for AI regulation, views the Anthropic crisis as a direct consequence of the industry's collective resistance to oversight. Tegmark, who helped organize an open letter calling for a pause in advanced AI development, asserts that companies like Anthropic have sowed the seeds of their own predicament by consistently lobbying against binding regulation, opting instead for a self-governance model.

Tegmark highlights a stark contradiction in the "safety-first" identity many AI companies project. Despite marketing themselves as champions of safety, firms such as Anthropic, OpenAI, Google DeepMind, and xAI have, according to Tegmark, consistently avoided supporting mandatory safety regulations. He points out that all four companies have recently retracted or weakened their internal safety commitments. Google dropped its "Don't be evil" and subsequent harm-prevention pledges, OpenAI removed "safety" from its mission statement, xAI disbanded its safety team, and Anthropic abandoned its promise not to release powerful AI systems until confident they wouldn't cause harm. Tegmark cynically notes the irony: these companies successfully lobbied for a regulatory vacuum, leaving them vulnerable when the government demands uses they object to, as there is "less regulation on AI systems in America than on sandwiches."

The common counter-argument from AI lobbyists, often citing a "race with China" as a justification for minimal regulation, is thoroughly debunked by Tegmark. He argues that this narrative, used to oppose any proposed regulation, ignores China's own proactive approach, such as considering bans on anthropomorphic AI to protect its youth. Tegmark posits that neither the Chinese Communist Party nor the U.S. government would tolerate an uncontrolled superintelligence developed by their own companies that could potentially overthrow them. He frames superintelligence as a fundamental national security threat, not a strategic asset to be rushed into existence in a competitive race.

Drawing an analogy to the Cold War, Tegmark explains that the U.S. won the economic and military dominance race against the Soviet Union without engaging in a suicidal nuclear arms race. He applies this logic to AI, asserting that the pursuit of uncontrollable superintelligence is akin to putting "nuclear craters in the other superpower" – a no-win scenario that would result in humanity losing control of Earth to "alien machines." He believes that once national security communities fully grasp this perspective, they will recognize uncontrollable superintelligence as a threat, not a tool.

The pace of AI development further underscores the urgency of these discussions. Tegmark notes that predictions from six years ago, which estimated decades until AI mastered human-level language and knowledge, have been proven wrong, with current AI systems rapidly progressing from high school to university professor levels in some areas. He cites a recent paper defining AGI, which showed GPT-4 at 27% and GPT-5 at 57% of the way there, suggesting that AGI might not be far off. He warns his MIT students that even a four-year timeline to AGI could mean job scarcity upon graduation, emphasizing the need for immediate preparation.

In the wake of Anthropic's blacklisting, the reactions of other major AI players are being closely watched. While OpenAI's Sam Altman publicly expressed solidarity with Anthropic, stating similar "red lines," Google has maintained silence, which Tegmark finds "incredibly embarrassing." xAI's stance also remains unknown. This moment, according to Tegmark, compels every major AI company to "show their true colors" regarding their commitment to ethical AI deployment and regulation.

Despite the current trajectory, Tegmark expresses a strange optimism, envisioning a clear alternative: treating AI companies like any other industry, subject to binding regulation. He suggests implementing "clinical trials" for powerful AI systems, requiring independent expert verification of their control mechanisms before release. This approach, he believes, could usher in a "golden age" of beneficial AI, free from existential risks, a path that, while not currently taken, remains achievable.