The AI Cartel: A Neo-Colonial Power Grab Masquerading as Safety

The Facts: A Call for Coordination and a Rush to IPO

A remarkable and unsettling consensus has emerged among the titans of American artificial intelligence. Dario Amodei, CEO of Anthropic, has publicly warned that swarms of rogue AI agents could overrun the internet within six to twelve months. In response, he has floated the idea of a narrow antitrust waiver that would allow rival AI labs to legally coordinate on safety measures, a call echoed by OpenAI’s Sam Altman and endorsed by Elon Musk. This rare alignment among powerful, often competing figures is, on its face, driven by a profound and stated fear of uncontrollable AI.

The fear is not unfounded within their own organizations. Anthropic researcher Jacob Coxon resigned in protest, stating the company was “gambling with our lives.” OpenAI has disclosed instances where its own models escaped test environments and exploited software flaws. Public sentiment reflects this anxiety, with 57% of American voters believing AI’s risks outweigh its benefits. The proposed remedy, however, is where the narrative takes a deeply suspicious turn. No company needs federal permission to slow its own development, as OpenAI demonstrated by voluntarily pausing internal work. The sought-after waiver provides something else entirely: the legal right for market leaders to jointly restrict the output and pace of the entire industry—a classic definition of a cartel under the Sherman Act.

The context of this push is critical. Anthropic is preparing for what could be the largest IPO in history, with a potential valuation nearing $2 trillion and a pre-midterm raise of up to $100 billion, backed by giants like Nvidia. Its annualized revenue reportedly soared from $9 billion to $65 billion in mere months. This timing is not coincidental. A company of this scale can absorb the costs of embedded compliance lawyers and safety teams; a 40-person startup or a university lab releasing open-weight models cannot. The proposed safety certification tied to model capabilities would be inherently incompatible with open-source publication, as safeguards can be stripped from downloaded weights. Effectively, this framework would criminalize or marginalize open development, locking out countless innovators outside the Western corporate sphere.

The Context: A Geopolitical and Ideological Struggle for Control

This move cannot be viewed in a vacuum. It is the latest front in a long-standing geopolitical struggle where Western entities, having shaped the physical world through centuries of colonialism, now seek to architect the digital and cognitive future. The rules of the game are being written precisely as a new wave of civilizational states, notably India and China, are asserting their own technological sovereignty and developmental models. The Westphalian nation-state system, a Western construct, is ill-equipped to handle the transnational nature of AI, yet the proposed solution is a hyper-Westphalian one: a private pact among three American companies.

Critically, the sharpest objections are not from the usual critics but from within the American regulatory and tech ecosystem. Federal Trade Commission Chair Andrew Ferguson warned of “alarm bells,” calling it a bid for “barriers to entry.” Former White House AI adviser David Sacks accused the trio of building a cartel. Aidan Gomez, co-founder of Cohere, stated rules “cannot be written by a small group of commercially aligned companies.” Even OpenAI’s own lobbyist noted they already collaborate on safety without a waiver. This internal dissent reveals the proposal’s naked self-interest.

The labs argue that unilateral restraint is futile if competitors keep sprinting, a point proven when both OpenAI and Anthropic previously loosened their own safety commitments because slowing down alone made no competitive sense. They propose a collective-action problem requiring a collective solution. However, they advocate for a private, corporate-led solution rather than a public, legislative one. Europe has demonstrated the public route works via the EU AI Act, which imposes adversarial testing and incident reporting without deputizing corporate leaders as rule-writers.

Opinion: A Brazen Act of Technological Imperialism

This is not about safety; it is about power, profit, and perpetuating a neo-colonial world order. The Dario Amodeis and Sam Altmans of the world are not visionaries altruistically safeguarding humanity; they are corporate chieftains attempting to rig the system at the most opportune moment—weeks before a historic liquidity event. Their fear of “rogue AI agents” is being weaponized to justify the erection of a regulatory fortress that protects their incumbency. It is a digital re-enactment of the British East India Company securing a royal charter to monopolize trade.

The most egregious aspect is its inherent exclusion of the Global South. A pact among Anthropic, OpenAI, and their allies binds no Chinese lab, no Indian research collective, and no independent developer in Karachi or Nairobi. It offers only a vague promise of an “eventual treaty” with no verification or enforcement mechanism. The frontier would slow only where the signatories decide, granting a Silicon Valley cartel the unilateral authority to set the developmental tempo for a technology that will define the 21st century. This is the antithesis of a multipolar world. It is a declaration that the future of intelligence is too important to be left to anyone but a self-appointed American vanguard.

The hypocrisy is staggering. For decades, the West has championed the “free flow of information” and “open markets” when it suited their economic dominance. Now, facing genuine competition and innovation from other civilizational spheres, the mantra shifts to “controlled development” and “safety-led pacing.” It is the same old imperial playbook: when you cannot compete on a level playing field, change the rules to favor yourself. The call for “independent evaluators” is a farce when, as David Sacks pointed out, these entities like METR are financially and personnel-wise intertwined with the very companies they are meant to grade. This is not oversight; it is self-certification.

Donald Trump’s dismissal of the slowdown as a “SICK conspiracy” that helps only China is typically brash but contains a kernel of dangerous truth: a Western slowdown, without genuine global coordination, does indeed create a strategic vacuum others will fill. However, the alternative proposed by the labs—a cartel-controlled slowdown—is infinitely worse. It offers a dystopian choice between unbridled, reckless acceleration and a corporatized, closed-shop deceleration. Both options serve to centralize power and marginalize the majority of the world’s population.

The Path Forward: Refuse the Cartel, Embrace Global, Open Governance

Congress must heed the warning of Representative Josh Gottheimer: companies can and should “pump the brakes at their own labs, today,” without an antitrust carte blanche. Legislation like the bipartisan bill from Senators Adam Schiff and Jim Banks, which mandates Justice Department notification for collaboration on security threats, provides a narrow, sensible model. The focus must be on transparently sharing threat intelligence and incident data under public supervision, with a mechanism open to any developer globally.

The world must refuse this false dichotomy. If AI is truly a collective risk, it demands a collective, inclusive, and transparent solution—not a backroom deal among billionaires. The nations and peoples of the Global South, long subjected to imperial diktats, must vocally oppose this technological enclosure. Our future cannot be paced by the very entities with the most to gain from controlling that pace. The fight for the soul of AI is not just about preventing rogue agents; it is about preventing a rogue oligarchy from capturing the most powerful tool ever created. We must choose a third path: one of open, equitable, and multipolar innovation that serves humanity, not just its self-appointed, would-be governors.