Is the future of warfare now written by algorithm? That’s the question Silicon Valley isn’t asking loudly enough as OpenAI, once a darling of the “AI for good” crowd, fully embraces its role as a Pentagon contractor. The real story here isn't OpenAI’s stated commitment to responsible AI – it’s the quiet erosion of ethical boundaries in the pursuit of profit and geopolitical advantage, and what that means for the rest of us. Just over two weeks ago, the company finalized a deal allowing the Department of Defense access to its AI in classified environments, a move met with surprisingly little public outcry given the initial promises of a different path.
Sam Altman insists the military won’t be using OpenAI’s technology to build autonomous weapons, but the agreement itself merely requires the Pentagon to adhere to its own guidelines – guidelines that are, frankly, permissive enough to drive a tank through. This isn’t a restriction; it’s a rubber stamp. OpenAI’s claim that the agreement prevents domestic surveillance feels equally flimsy, lacking any concrete mechanism for enforcement. The speed of this pivot is what’s truly unsettling. OpenAI isn’t simply entering a market it once shunned; it’s sprinting towards it, fueled by the massive costs of AI training and a relentless hunt for new revenue streams – including, reportedly, advertising. Or perhaps, as Altman often frames it, it’s a matter of national security, a necessity for liberal democracies to compete with China in the AI arms race.
Drawn from technologyreview.com.
The immediate implications are playing out against a backdrop of escalating tensions. OpenAI’s technology is poised to integrate with existing military tools, a process mirroring that of Elon Musk’s xAI and its Grok model. But the integration isn’t seamless. The Pentagon is still grappling with the fallout from Anthropic’s refusal to allow its AI to be used for “any lawful use,” a stance that led to President Trump ordering the military to stop using it and designating Anthropic a supply chain risk – a designation the company is currently fighting in court. If the conflict with Iran continues to escalate, what will OpenAI’s role be? A defense official recently described a scenario where AI prioritizes potential targets based on logistical data, intelligence analysis, and a deluge of information from various sources. A human analyst would then “double-check” the AI’s recommendations.
But let’s be real: if a human is truly vetting every AI-generated target, what’s the point of the speed boost? For years, the military has utilized AI systems like Maven to analyze drone footage and identify potential targets. OpenAI’s models, like Anthropic’s Claude, will likely function as a conversational layer on top of these existing systems, allowing users to ask for interpretations and recommendations. This isn’t incremental improvement; it’s a fundamental shift in how warfare is conducted, testing the limits of generative AI in the field for the first time in a live conflict. The stakes aren’t just about military efficiency; they’re about accountability and the potential for unintended consequences.
Beyond targeting, OpenAI is also deepening its ties with companies like Anduril, a major player in drone and counter-drone technology. A $20 billion contract with the US Army will integrate Anduril’s Lattice interface – which controls everything from drone defenses to autonomous submarines – with legacy military equipment, layering AI on top of it all. If OpenAI’s models prove useful, they’ll be seamlessly incorporated into this broader warfare stack. Even seemingly mundane applications are being targeted. Defense Secretary Pete Hegseth is pushing for the widespread adoption of AI tools like GenAI.mil, which provides secure access to commercial AI models for tasks like drafting policy documents and managing contracts. Google’s Gemini and xAI’s Grok are already on the platform, despite Grok’s history of generating problematic content.
OpenAI’s addition to GenAI.mil isn’t about influencing critical decisions in Iran; it’s about establishing an “all-in” attitude towards AI across the Pentagon, transforming every aspect of warfare – from targeting to paperwork. The message is clear: AI is no longer a futuristic concept; it’s an integral part of the US military’s present and future. And OpenAI is rapidly securing its place at the center of it all. But here’s what everyone is missing: the normalization of this integration. The public isn’t debating if AI should be used in warfare, but how.
Expect to see a surge in “AI safety” initiatives designed to manage the risks, not prevent them. The real question isn’t whether OpenAI can build ethical AI for the military – it’s whether the military will allow it to. Watch closely for the first publicly acknowledged incident where an AI-driven targeting recommendation leads to civilian casualties. That moment will expose the hollowness of the current safeguards and force a reckoning with the true cost of this technological embrace.
