OpenAI Faces Backlash Over Pentagon Deal Despite Added Surveillance Restrictions
Key Takeaways
- ▸ChatGPT uninstalls reportedly surged 300% after OpenAI announced its Pentagon partnership, prompting CEO Sam Altman to acknowledge the deal was "opportunistic and sloppy"
- ▸OpenAI amended its contract to prohibit "intentional" domestic surveillance of U.S. persons, but critics argue the vague language provides little actual protection
- ▸The Electronic Frontier Foundation warns that terms like "deliberate," "intentionally," and "unconstrained" are "weasel words" that allow the government to maintain broad surveillance capabilities while appearing to accept restrictions
Summary
OpenAI is facing widespread criticism following its decision to partner with the U.S. Department of Defense after rival Anthropic refused to remove restrictions against using AI for surveillance and autonomous weapons systems. The controversy intensified when early reports showed ChatGPT uninstalls rose nearly 300% after the deal was announced, prompting protests from both users and employees. CEO Sam Altman acknowledged the initial agreement was "opportunistic and sloppy" and attempted to clarify the terms by republishing an internal memo stating the AI system would not be "intentionally used for domestic surveillance of U.S. persons and nationals" consistent with applicable laws including the Fourth Amendment.
However, the Electronic Frontier Foundation (EFF) has raised serious concerns about the vague language in the amended contract, particularly terms like "intentionally," "deliberate," and "unconstrained." The EFF argues that the U.S. government has historically embraced lax interpretations of surveillance laws that enable mass data collection while claiming compliance with legal frameworks. The organization points out that intelligence and law enforcement agencies often rely on "incidental" collection or commercially purchased data to circumvent stronger privacy protections, making OpenAI's restrictions potentially meaningless in practice.
The amended contract includes provisions stating the AI system shall not be used for "deliberate tracking, surveillance, or monitoring of U.S. persons" or "unconstrained monitoring of U.S. persons' private information." Critics characterize these as "weasel words" that create ambiguity and protect parties from accountability while appearing to establish meaningful limits. The EFF's analysis suggests that similar to negotiations with Anthropic where the Pentagon agreed to adhere to restrictions only "as appropriate," the government may be publicly committing to principles while retaining broad operational flexibility.
- The deal highlights a competitive dynamic in the AI industry, with OpenAI filling the gap left by Anthropic's refusal to remove restrictions on surveillance and autonomous weapons use
