What Happened
As OpenAI's deal with the Department of Defense drew public scrutiny in late February and early March 2026, TechCrunch identified a broader structural problem: neither OpenAI, Anthropic, nor the US government has established a coherent framework for how AI companies should engage with military and national security clients. OpenAI CEO Sam Altman acknowledged the deal was "definitely rushed" and that "the optics don't look good." The situation played out through a series of escalating events - Anthropic refusing terms, Trump directing agencies to drop its products, the DoD designating Anthropic a supply-chain risk, and then OpenAI announcing its own deal.
Why It Matters
The absence of a framework has real consequences for everyone involved. Anthropic held two specific red lines - no mass surveillance of US persons, no autonomous lethal decision-making - and was designated a supply-chain risk for holding them. OpenAI signed broader terms and faced a consumer backlash measured in hundreds of thousands of uninstalls. Neither outcome reflects a functioning procurement system.
For AI companies building enterprise and government businesses, the current environment is genuinely uncertain. The rules are being negotiated deal-by-deal, with political pressure from the executive branch now a direct variable in contract terms. That's a difficult environment for product planning, ethics policy writing, and investor communication. A company can have a published acceptable use policy and still find that policy becomes grounds for vendor blacklisting.
Defense contractors have spent decades building compliance structures, cleared personnel pipelines, and government relations capabilities. AI companies are attempting to fast-track that maturation while simultaneously managing consumer brand perception - two priorities that are now in direct conflict in ways the companies haven't resolved.
For buyers in regulated industries that use AI providers, this creates a new category of supply chain risk: what happens to your AI tools if your vendor refuses a government directive?
Our Take
This is less a story about OpenAI or Anthropic specifically and more about an industry that scaled faster than its governance structures could follow. The companies that work out explicit, public policies for government engagement now will be better positioned than those still improvising deal-by-deal. The absence of a framework isn't just a PR problem - it's a business risk that will affect enterprise sales cycles as government-adjacent buyers ask harder questions about AI vendor reliability.
The practical fix is not simple. Creating a framework requires coordinating between AI companies with different values and business models, multiple government agencies with different risk tolerances, and Congress, which has shown limited appetite for AI legislation. What the OpenAI-Anthropic episode does is create a documented case study that makes the costs of the current absence of framework concrete. That may be the most important short-term contribution to eventually getting one in place.