OpenAI Violated Catastrophic Primary evidence Military

OpenAI: usage policy bars 'military and warfare' applications

Made 2023-03-23 00:00:00+00 · Audience · General public

// The promise

“Disallowed usage of our models: … Activity that has high risk of physical harm, including: Weapons development; Military and warfare.”

OpenAI Usage Policies, archived 8 January 2024 ↗

For nine months OpenAI's published usage policy explicitly banned military and warfare applications. In January 2024 the prohibition was quietly removed and within days the company confirmed an active working relationship with the United States Department of Defense.

// What actually happened (2 sourced violations)

#1 2024-01-12 00:00:00+00 295 days later primary evidence

OpenAI removes the explicit 'military and warfare' prohibition from its usage policy.

The Intercept, 12 January 2024 ↗
#2 2024-12-04 00:00:00+00 622 days later primary evidence

OpenAI confirms cybersecurity work with the US Department of Defense; later announces partnership with weapons-maker Anduril.

Reuters, 4 December 2024 ↗

PROVE IT How verification works →