OpenAI’s Slippery Military Policy
OpenAI’s CEO, Sam Altman, is taking heat for a military deal that bypassed their own no-military-use policy. Employees aren’t thrilled, especially after Anthropic’s $200 million Pentagon deal fizzled. Altman admitted the situation was ‘sloppy’—a PR move straight out of the ‘Oops, my bad’ playbook.
Despite OpenAI’s 2023 policy banning military use, the Pentagon was already tinkering with Azure OpenAI, courtesy of Microsoft. OpenAI employees were left scratching their heads, wondering if Microsoft’s military-friendly moves were a loophole or just business as usual.
Microsoft’s Convenient Loophole
Microsoft, OpenAI’s sugar daddy, has been cozy with the Department of Defense for ages. Their Azure OpenAI service waltzed into government hands in 2023, not bound by OpenAI’s policies. Frank Shaw, Microsoft’s spokesperson, confirmed this, sidestepping specifics on Pentagon timelines.
OpenAI employees noticed Pentagon folks roaming their San Francisco offices, sparking confusion and concern. The question was simple: Did OpenAI’s ban apply to Microsoft? Apparently not, as Azure OpenAI danced around the rules, leaving employees in the dark.
Policy Changes and New Partners
By January 2024, OpenAI quietly axed its military ban, a fact employees discovered via The Intercept. Leadership later addressed this at an all-hands meeting, assuring them of careful future steps. Meanwhile, OpenAI partnered with Anduril for ‘national security missions,’ only for unclassified work, they claimed.
Palantir also came knocking in 2024 with a high-risk proposal, which OpenAI declined. Yet, they still found ways to collaborate with Palantir. Employees, skeptical of their models’ reliability, voiced concerns in a public Slack channel, questioning the wisdom of military partnerships.
Quick Facts
- •💡 OpenAI banned military use of its AI in 2023.
- •💡 Pentagon tested OpenAI models via Microsoft in 2023.
- •💡 OpenAI’s policy loophole allowed Pentagon access.
- •💡 Microsoft’s Azure OpenAI not bound by OpenAI’s policies.
- •💡 OpenAI updated policies in 2024, allowing military use.