Commentary: Altman's Opaque AI Is creating a new security dilemma

The OpenAI CEO Sam Altman has sparked a dangerous AI arms race by unveiling a new model called Astra, which reportedly uses a method called "recurrent depth." This method makes reasoning more efficient by not spelling it out in human language, potentially making it harder for researchers to monitor the system for harmful behavior.

Despite OpenAI's prior pledges to prioritize safety, the company appears to be prioritizing speed and commercial advantage over transparency. Astra's use of recurrent depth decreases the company's ability to monitor its chain-of-thought, which can help researchers understand how the technology arrives at its decisions. This approach, while providing a commercial advantage, raises concerns about the potential for the technology to become more powerful without adequate oversight, leading to an escalation in the AI arms race and a security dilemma similar to historical conflicts over weapons development.