tech
March 12, 2026
OpenAI's Pentagon Deal Continues to Face Both Internal and External Scrutiny
Turn any article into a podcast. Upgrade now to start listening.
TL;DR
- OpenAI has signed a deal with the Department of Defense to provide technology for classified networks.
- OpenAI's robotics team lead resigned, citing concerns about the deal being "rushed without guardrails defined."
- Critics question if the deal adequately prevents mass surveillance and autonomous weapons, as claimed by OpenAI.
- The deal follows Anthropic's refusal to remove restrictions on autonomous weapons and mass surveillance, leading to its ban by the Trump administration.
- OpenAI CEO Sam Altman claims the Department of Defense agrees with safety principles against domestic mass surveillance and autonomous weapons.
- The public wording of OpenAI's agreement is seen by analysts as softer than Anthropic's, potentially allowing for 'lawful' mass data processing.
- National security laws may interpret 'surveillance' narrowly, allowing for the collection of data on U.S. citizens incidentally while targeting non-Americans.
- AI's ability to analyze large datasets makes previously difficult-to-use consumer data more accessible for government analysis.
- OpenAI has made changes to its contract language regarding the use of commercially acquired data for domestic surveillance.
- Skepticism exists about OpenAI's ability to uphold its safety restrictions against government pressure.
- The Trump administration's designation of Anthropic as a supply-chain risk demonstrates its willingness to use strong measures.
- OpenAI plans to use its technical 'safety stack' to monitor usage and prevent violations.
- Experts doubt the technical 'safety stack' is sufficient to prevent illegal use, citing the potential for 'jailbreaks' in AI systems.
- Some OpenAI employees have expressed concerns, with one calling the deal "rushed" and another suggesting independent legal counsel review.
- 100 OpenAI employees signed a public letter against domestic mass surveillance and autonomous weapons.
- OpenAI's policy on military sales has evolved from a ban to allowing sales since January 2024.
- Competition is intensifying between OpenAI and Anthropic, with Anthropic projected to potentially surpass OpenAI in revenue.
- The deal raises questions about who should determine the permissible uses of AI technology.
Continue reading the original article