OpenAI and the Pentagon: AI at the Heart of Tensions with Iran

OpenAI and the Pentagon: A Controversial Partnership

Just over two weeks ago, OpenAI signed an agreement that has sparked intense discussions. This agreement allows the Pentagon to leverage OpenAI's technology in classified environments. However, questions remain about the true scope of this agreement. Sam Altman, the CEO of OpenAI, stated that the military would not be able to use their technology to develop autonomous weapons. Yet, the agreement simply stipulates that the military must adhere to its own guidelines, which are relatively permissive regarding autonomous weapons. Furthermore, OpenAI's promise not to allow its technology to be used for domestic surveillance also seems uncertain.

OpenAI's motivations remain unclear. While it is not the first tech company to accept military contracts, the speed with which OpenAI has shifted its position is striking. It is possible that financial motivations are at play, as OpenAI is heavily investing in training its AI models and seeking to diversify its revenue streams, including through advertising. Alternatively, Sam Altman may genuinely believe that liberal democracies, and their armed forces, must have access to the most advanced AI technologies to effectively compete with China.

The Future of OpenAI's AI in Conflicts

The crucial question is how this technology will be used. OpenAI has decided to engage at the heart of the conflict at a time when the United States is ramping up its operations against Iran, with an increasing role for AI in these operations. Thus, the question arises: where will OpenAI's technology be deployed in this context, and what applications will be accepted by its clients and employees?

Integration and Potential Applications

Although the agreement with the Pentagon is in place, the integration of OpenAI's technology into classified environments is not yet effective. This integration requires compatibility with other tools already used by the military. For example, xAI, founded by Elon Musk, which has also signed an agreement with the Pentagon, must undergo a similar process for its AI model Grok. However, there is pressure to expedite this process, partly due to the controversies surrounding the technology used thus far. Anthropic, another AI company, has refused to allow its AI to be used for "any lawful purpose," which led President Trump to order its use by the military to be halted. The Pentagon has even classified Anthropic as a supply chain risk, a decision the company is currently contesting in court.

If the conflict in Iran persists until the integration of OpenAI's technology, what could it be used for? According to a defense official, a human analyst could use the AI model to analyze a list of potential targets and determine which ones to prioritize for strikes. The model could take into account logistical information, such as the location of aircraft or supplies, and analyze various types of data, including texts, images, and videos.

A human would be tasked with manually verifying these results, according to the official. However, this raises a question: if human verification is necessary, how does AI actually accelerate targeting and strike decisions?

AI in Service of Defense

For several years, the military has been using an AI system called Maven for tasks such as automatic drone image analysis to identify potential targets. It is likely that OpenAI's models, like Claude from Anthropic, will offer a conversational interface that allows users to request intelligence interpretations and recommendations on which targets to strike first.

The use of generative AI to advise on actions to take on the ground is a new development being tested in Iran. In late 2024, OpenAI announced a partnership with Anduril, a company specializing in the manufacture of drones and counter-drone technologies for the military. This agreement stipulates that OpenAI will collaborate with Anduril to conduct real-time analyses of drones attacking U.S. forces and assist in neutralizing them. An OpenAI spokesperson clarified that this does not violate the company's policies, which prohibit "systems designed to harm others," as the technology is used to target drones rather than people.

Anduril provides a range of counter-drone technologies to military bases worldwide, although the company has not specified whether its systems are deployed near Iran. Neither company has provided an update on the project's progress since its announcement. However, Anduril has long trained its own AI models to analyze camera images and sensor data to identify threats. What it focuses less on are conversational AI systems that allow soldiers to directly query these systems or receive advice in natural language, an area where OpenAI's models could fit in.

Military Stakes

The stakes are high. On March 1, six members of the U.S. forces were killed in Kuwait during an Iranian drone attack that was not intercepted by U.S. air defenses.

Anduril's interface, called Lattice, serves as the control point for soldiers, allowing them to manage everything from drone defenses to missiles and autonomous submarines. The company recently secured massive contracts totaling $20 billion from the U.S. military to integrate its systems with existing military equipment and add a layer of AI. If OpenAI's models prove useful to Anduril, Lattice is designed to quickly integrate them into this broader military infrastructure.

The Impact of AI on Military Administration

In December, Defense Secretary Pete Hegseth encouraged millions of personnel in administrative roles within the military—such as contracts, logistics, and procurement—to use a new AI tool. Named GenAI.mil, this tool allows staff to securely access commercial AI models to perform the same types of tasks as in the private sector.

Google Gemini was one of the first models available. In January, the Pentagon announced that Grok from xAI would also be added to the GenAI.mil platform, despite incidents where the model had generated antisemitic content and non-consensual deepfakes. OpenAI followed in February, announcing that its models would be used to draft policy documents and contracts, and to assist with administrative support for missions.

While the use of ChatGPT for unclassified tasks on this platform is unlikely to directly influence sensitive decisions in Iran, OpenAI's presence on the platform is significant in another way. It reflects the Pentagon's commitment to AI, a commitment that Hegseth has consistently promoted. The message is clear: AI is transforming every aspect of how the United States conducts its military operations, from targeting decisions to administrative management. And OpenAI is increasingly establishing itself as a key player in this transformation.