OpenAI reveals contract terms and boundaries with Department of War

According to OpenAI, its contract with the Department of War is "superior" and incorporates more safety measures than the one Anthropic refused, leading to its blacklisting.

In a publication made on Saturday, OpenAI revealed some of the terms from its agreement with the Department of War, highlighting sections that ensure its technology won’t be used for extensive domestic monitoring or in the operation of independent weapons or high-stakes decision systems like "social credit" assessments.

"We believe our agreement offers more protections than any prior arrangement for classified AI deployments, including Anthropic's," OpenAI's blog mention stated.

"Our agreement incorporates a comprehensive, multi-layered approach to safeguard our red lines. We maintain full control over our safety protocols, deploy through the cloud, involve authorised OpenAI staff, and have strong contractual defenses, all in addition to the existing robust protections in US. law."

OpenAI's CEO, Sam Altman, addressed inquiries on social media soon after posting the company's blog, answering questions about the nature of OpenAI's government contract.

In a session mimicking an Ask-Me-Anything, he reiterated OpenAI's contract being superior to Anthropic's, highlighting benefits not only for the AI industry but also for American citizens.

"Anthropic appeared more absorbed in specific prohibitions outlined in the contract, rather than referring to applicable laws, which we found agreeable," Altman explained when questioned about OpenAI's willingness to partner with the government unlike its competitor. "I suspect Anthropic may have desired more operational authority than we had."

The contract between OpenAI and the federal government succeeded Anthropic being blacklisted and classified as a supply chain risk due to its refusal to comply with the military’s usage criteria for the Claude model.