OpenAI Staff Fume Over Pentagon Deal
Some OpenAI staff are reportedly upset about the company's contract with the Pentagon. The internal backlash highlights the ongoing cultural tension within leading AI labs about the application of their technology in government and defense.
OpenAI's deal with the Pentagon follows a key policy change in January 2024, when the company quietly removed language from its usage policy that explicitly banned "military and warfare" applications. The updated policy still forbids using its AI to "develop or use weapons," but the broader prohibition on military use was eliminated. The agreement was announced just hours after the Pentagon dropped a deal with rival AI firm Anthropic. Anthropic had refused Pentagon demands to remove safeguards that would prevent its AI from being used for mass surveillance or fully autonomous weapons systems. Following this refusal, the Trump administration designated Anthropic a "supply chain risk." As part of a broader initiative, OpenAI, Google, Anthropic, and xAI are positioned to receive contracts with a ceiling of $200 million each to help the Department of Defense integrate advanced AI. OpenAI's specific role involves deploying its advanced AI systems within the Pentagon's classified networks. The deal triggered significant backlash both internally and externally. An open letter signed by nearly 100 OpenAI employees and almost 800 Google staffers urged companies to reject demands for using AI in surveillance or autonomous weapons. A public "delete ChatGPT" campaign also gained traction online, with uninstalls of the app reportedly surging. In response to the criticism, CEO Sam Altman acknowledged the rollout of the agreement appeared "opportunistic and sloppy" and was "rushed." Following the backlash, OpenAI amended the contract to explicitly prohibit the use of its systems for intentional domestic surveillance of U.S. citizens and by intelligence agencies like the NSA without further modifications. Despite the revisions, CEO Sam Altman reportedly told employees in an all-hands meeting that OpenAI cannot ultimately control how the U.S. military uses its technology once it has been deployed. This stance highlights the ongoing tension between creating powerful AI and governing its use in sensitive military applications.
