OpenAI CEO Sam Altman is revising the company's controversial Pentagon AI contract after intense public backlash. The amended deal now explicitly prohibits using OpenAI technology for "domestic surveillance of U.S. persons and nationals" and extends restrictions to "commercially acquired" personal data — a stricter standard than the original language covering only private information.
Altman acknowledged the misstep in an employee memo, stating he "regretted moving so fast" and called the rollout "opportunistic and sloppy." Intelligence agencies like the NSA are now explicitly excluded from the deal, with any future use requiring a separate contract modification. The backlash was so severe that rival Anthropic — which was banned from federal agencies for refusing weapons work — briefly overtook ChatGPT as the top App Store download.
The revision marks a rare public retreat for OpenAI, which had moved quickly to secure the Pentagon contract after Anthropic's blacklisting left a vacuum in government AI services. Critics argue the guardrails still lack independent oversight, while supporters say the amendments represent meaningful protections against misuse.
