Abstract:
A knife is neither good nor bad. Its moral weight is entirely determined by whose hand holds it. This principle, elementary to classical Islamic jurisprudence, is also the only honest framework for evaluating artificial intelligence, a technology whose advocates speak of potential and whose critics speak of risk, while the foundational question goes unasked: who already controls it, and how was that control established? This article documents that governance of advanced AI has been decided not through public deliberation or democratic oversight, but through the same mechanisms that have historically concentrated every catastrophic technology in the hands of the powerful: revolving doors, the subordination of ethics to financial pressure, and the exclusion of any institution that might impose a constraint. OpenAI removed its prohibition on military use without announcement. Sam Altman signed a Pentagon deal on the day the war on Iran began, then told employees that once the technology is handed over, the company no longer controls its use. Palantir became the default operating system of the American military through ten billion dollars in contracts. The knife has already been placed in specific hands. The debate about who should hold it is occurring after that decision was made.
Main Article: A principle in Islamic jurisprudence holds that instruments are morally neutral. A knife does not carry the weight of what it has done or might do. A sword, a poison, a fire, a law, a technology: none of these carries inherent good or evil. What determines their moral character is the hand that holds them, the intention that directs the hand, and the system that puts the hand in a position to hold them at all. This principle did not originate as a commentary on technology. It originated as a framework for evaluating action, responsibility, and the distribution of power in a world where the same instrument serves radically different purposes depending on who commands it.
It is also the only intellectually honest framework for evaluating artificial intelligence in the specific historical moment in which the technology has arrived.
The public debate about AI has been conducted almost entirely in two registers. In the first, advocates for the technology describe its potential to cure disease, accelerate scientific discovery, reduce poverty, and expand human capability. In the second, critics describe its potential to eliminate employment, produce autonomous weapons, enable mass surveillance, and accelerate the concentration of power in entities that are already the most powerful on Earth. Both registers are engaged with the instrument. Neither is engaged with the foundational question: who already controls it, and through what process was that control established?
The question matters because the answer has already been given, not in a policy document or an ethics framework or an international treaty, but through the institutional mechanisms that have historically determined who holds every catastrophic technology humanity has produced. [1]
On January 10, 2024, OpenAI quietly revised its usage policies, removing language that had explicitly prohibited the use of its AI systems for military and warfare purposes. The revision was not announced in a press release. It was not accompanied by a public explanation of the reasoning. Users who had understood the prohibition to be in effect were not notified that it had been removed. [2] In the months that followed, the company hired a former deputy chief digital and artificial intelligence officer from the Office of the Secretary of Defense, a former Biden National Security Council spokesperson, a former assistant secretary of defense for Indo-Pacific affairs, and multiple former Senate aides with active lobbying registrations on behalf of the company. [2] The architecture of the revolving door between the company and the institutions it was preparing to supply was assembled before the supply relationship became public.
On February 27, 2026, the day before the United States and Israel launched their military campaign against Iran, Sam Altman announced on X that OpenAI had reached a deal with the Department of Defense, framing it as a patriotic necessity ensuring that democratic AI maintains safeguards that adversary systems lack. [3] The contract was classified. Its full terms were not released. When The Intercept requested the specific contract language establishing the assurances Altman described, OpenAI did not respond. [4] When employees asked at an all-hands meeting what the company would do if the military used its technology in ways that crossed the ethical lines Altman had publicly drawn, he told them that once the technology is handed to the government, operational decisions belong to the government. The company builds it. It does not govern its use. [5]
Altman had previously written publicly that his greatest fear about Donald Trump was war. He signed the Pentagon deal on the day the war began.
Across the same period, Palantir Technologies, founded in 2003 with seed funding from the Central Intelligence Agency investment arm In-Q-Tel, became the default operating system of the American military. By August 2025, the United States Army had consolidated 75 separate contracts with Palantir into a single enterprise agreement worth up to ten billion dollars over a decade, effectively making the company the vendor whose software the military cannot fight without. [6] By March 2026, the Deputy Secretary of Defense signed a memo formally designating the Maven Smart System of Palantir, the AI platform that analyzes battlefield data and identifies targets, as a program of record across the entire joint force, locking it into long-term institutional status and moving its oversight into the Chief Digital Artificial Intelligence Office of the Pentagonl Intelligence Office. [7] Maven had already been deployed as the primary decision support system for United States Central Command during the war on Iran, integrated with other AI components to process targeting data at a scale and speed that no human analytical chain could replicate. [1]
This is the knife, and these are the hands that hold it.
The people whose decisions shaped this outcome are not anonymous bureaucrats operating within systems too complex to trace. They are documented, named, and their incentive structures are not difficult to understand. Peter Thiel, a co-founder of Palantir, spent years cultivating relationships in Washington that turned his company from a CIA-funded intelligence contractor into the institutional backbone of American military AI. Elon Musk, whose company xAI is developing its own large language models and whose influence over American government has expanded into a formally constituted advisory structure, accumulated his fortune and his proximity to power through a combination of government subsidies, aggressive labor practices, and a documented willingness to use his platforms to shape political outcomes in his preferred direction. Sam Altman runs a company that was founded as a nonprofit committed to ensuring artificial general intelligence benefits all of humanity, revised its policies against military use without telling its users, signed a classified weapons deal on the first day of a war, and then told his employees that he cannot govern what happens to his technology once it leaves his hands. [2]
These are not people selected by any democratic process to make decisions about the most consequential technology in human history. They are people who accumulated sufficient capital and proximity to power, through mechanisms that included lobbying, revolving-door hiring, political relationship cultivation, and the systematic subordination of stated ethical commitments to financial incentive, to position themselves as the entities whose judgment the technology reflects by default. [2]
The nuclear parallel is not decorative. The chapter in this publication on Israeli nuclear governance documented that a state can hold between 90 and 400 nuclear warheads while being formally exempt from every international oversight obligation, because it never admitted the weapons exist. The same structural principle operates in advanced AI governance: the most consequential deployment decisions are being made in classified contracts whose terms are not released, by companies whose internal deliberations are not subject to public oversight, at a speed that outpaces the institutional capacity of any regulatory body to respond. The argument made by Altman and echoed across Silicon Valley is that American AI companies must engage with the military because if they do not, adversary nations will develop military AI without any ethical safeguards at all. The argument is structurally identical to the one that has justified every arms race in modern history. It has the same logical shape as the nuclear deterrence argument. It produces the same result: the fastest possible acceleration of the most dangerous capabilities, justified at each step by the imperative of not allowing the adversary to hold them alone. [3]
The people making this argument are not disinterested analysts of geopolitical risk. They are the owners of companies whose valuations increase with every government contract awarded and every ethical constraint removed. The stock price of Palantir more than doubled in the year following its Army enterprise agreement. The financial incentive to frame military AI deployment as a moral necessity rather than a commercial calculation is not incidental to the argument. It is its structural foundation. [6]
What makes the present moment different from prior concentrations of dangerous technological power is not the danger itself but the speed. Nuclear weapons required the resources of nation states and years of industrial mobilization to develop. The geographic and logistical constraints on their spread were significant. Advanced AI systems are being developed by private companies operating under private governance, deployed through classified contracts, integrated into military targeting chains faster than any existing legal or institutional framework can track, and scaled by the same computational infrastructure that serves consumer products, commercial enterprise, and weapons systems simultaneously. The knife is the same knife in every hand. The hand it ends up in is determined by who can afford the handle.
The Islamic jurisprudential principle that opens this analysis does not end with the observation that instruments are neutral. It continues: because instruments are neutral, the moral responsibility for their use falls entirely on the system that determines whose hand holds them. A society that allows catastrophic instruments to accumulate in the hands of those who accumulated the most capital through the most aggressive means has not left the question of the knife unanswered. It has answered it by default, through the absence of any system designed to give a different answer.
What humanity needs is not a better argument about the potential of AI. It does not need another ethics framework published by the companies that will violate it when a sufficiently lucrative contract arrives. It needs what it has always needed when a new instrument of catastrophic potential appears: a governance system whose design begins with the question of whose hand should hold it, rather than discovering whose hand already does after the contracts are signed and the classified systems are deployed. The knife was handed over before the public debate began. The question now is whether the institutions capable of taking it back exist, or whether they will need to be built before the people currently holding it decide there is nothing left to discuss.
References:
1. "Algorithm of War: How Palantir Became the Pentagon's Indispensable AI Arsenal in Wars Abroad." Press TV, April 26, 2026.
2. "OpenAI Is Bleeding Cash. Its Solution? Military Contracts." Jacobin, April 5, 2026.
3. "OpenAI Strikes Deal with Pentagon to Use Tech in Classified Network." Al Jazeera, February 28, 2026.
4. "OpenAI on Surveillance and Autonomous Killings: You Are Going to Have to Trust Us." The Intercept, March 8, 2026.
5. "Altman Tells Staff: OpenAI Cannot Control Military Use." TechBuzz AI, March 4, 2026.
6. "Palantir Lands $10 Billion Army Software and Data Contract." CNBC, August 1, 2025.
7. "Pentagon to Adopt Palantir AI as Core US Military System, Memo Says." Reuters, March 20, 2026.