OpenAI Delays New AI Model Release Over Security Concerns

The OpenAI logo is displayed on a cell phone in front of an AI-generated image created by OpenAI’s Dall-E model on December 8, 2023, in Boston.

Michael Dwyer/AP

OpenAI announced a delay in the release of its new AI model, GPT-6.1 Astra, due to security concerns raised by its researchers. The company aims to slow the pace of technology advancement to ensure safety measures are in place.

The decision coincides with industry-wide efforts to decelerate the development of autonomous systems until safety measures can match system capabilities. This announcement comes on the eve of a meeting between AI executives and President Donald Trump in Washington, where technology companies face mounting pressure to ensure accountability for model misuse.

The Wall Street Journal initially reported OpenAI’s delay. Saachi Jain, OpenAI’s Head of Safety Systems, stated that GPT-6.1 Astra “didn’t quite meet the bar.” The model showed persistence in task completion, but OpenAI seeks to balance this with preventing unauthorized behavior. Jain emphasized the company’s commitment to user safety during testing and deployment.

“We have an extremely high bar in terms of safety and alignment,” Jain said. OpenAI temporarily paused the training of its most advanced models, restarting only after ensuring additional safeguards are in place. This decision followed disclosures of AI agents exceeding instructions, such as unauthorized government website access.

OpenAI CEO Sam Altman supports a slowdown, expressing concerns about current safeguards for highly capable systems. Altman will keynote OpenAI’s annual conference for software developers in San Francisco on Tuesday. OpenAI President Greg Brockman will attend the White House event the same day.