Sam Altman aims to 'help de-escalate' tensions with Pentagon as OpenAI employees support Anthropic

In a memo to OpenAI employees, CEO Sam Altman expressed his desire to assist in reducing tensions between the startup Anthropic and the Department of Defense (DoD). He emphasized the company's longstanding position against using artificial intelligence for mass surveillance or autonomous lethal weaponry, stating that human oversight is essential in high-stakes automated decisions. This clarion call for ethical boundaries was shared with staff late Thursday and aims to align OpenAI's principles with those of Anthropic. The pressure is mounting on Anthropic, which faces a deadline to decide whether to grant the DoD unrestricted access to its AI technology for all lawful applications. The startup is seeking guarantees that its models won't be utilized for fully autonomous weapons or intrusive domestic surveillance, a demand the DoD has yet to address. Altman's memo seeks to demonstrate OpenAI's solidarity with Anthropic on these critical issues. Before Altman’s communication, a number of OpenAI employees had already rallied in support of Anthropic, voicing their concerns on social media. Around 70 staff members signed an open letter titled "We Will Not Be Divided," aiming to foster unity and understanding amidst the pressures from the defense department. In a subsequent interview, Altman acknowledged the complexities of the situation while affirming his trust in Anthropic's commitment to safety. He noted that despite differences, he appreciates their support for military operations. OpenAI's relationship with the DoD has been significant, having secured a $200 million contract last year to utilize its models for non-classified applications. Looking ahead, Altman indicated that OpenAI is exploring possibilities for a deal with the DoD that aligns with its ethical standards, including implementing technical safeguards and deploying personnel to ensure compliance. He underscored the importance of making principled decisions over merely pursuing appearances of strength, recognizing the nuanced challenges involved. More discussions with OpenAI's safety teams are scheduled as the company navigates this evolving landscape.