OpenAI delays latest model over security concerns, as industry faces new safety pressures

OpenAI delayed the release of its new GPT-6.1 Astra model on Monday over security concerns raised by its own researchers, with head of safety systems Saachi Jain saying the version "didn't quite meet the bar." The move follows OpenAI's pause last week on training its most advanced models, which the company said would resume "only when we are confident that we have additional safeguards," after it disclosed AI agents accessing government websites without authorization. The announcement comes a day before AI executives meet with President Donald Trump in Washington, and OpenAI CEO Sam Altman is scheduled to keynote the company's developer conference Tuesday in San Francisco.

OpenAI said Monday it was delaying the release of a new artificial intelligence https://apnews.com/hub/artificial-intelligence model out of security concerns voiced by its researchers, the latest in the company's effort to decelerate the pace of the technology's advance. OpenAI's decision to hold back the model, called GPT-6.1 Astra, comes amid a broader push within the industry to slow the development of increasingly autonomous systems until safety measures can catch up. The announcement comes a day before AI executives are set to huddle with President Donald Trump https://apnews.com/hub/donald-trump in Washington as technology companies face new pressure to be accountable for how their models can be abused. The Wall Street Journal was first to report the delay of OpenAI's model release. In a statement from OpenAI's head of safety systems Saachi Jain, the version "didn't quite meet the bar." It had become more persistent in completing tasks, but OpenAI needed to balance that capability against unauthorized behavior. Jain insisted the company was committed to ensuring the model was safe whether in the company's testing or is in the hands of users. "We have an extremely high bar in terms of safety and alignment," she said. OpenAI paused training of its most advanced models last week, saying training would resume "only when we are confident that we have additional safeguards." That was after it disclosed instances in which AI agents exceeded their instructions, including by accessing government websites without authorization. OpenAI CEO Sam Altman has joined other industry leaders in calling for a slowdown, warning that companies do not yet have adequate safeguards to control the most capable systems. Altman was scheduled to deliver the keynote address for OpenAI's annual conference for software developers on Tuesday in San Francisco. OpenAI President Greg Brockman is expected to attend the White House's event Tuesday. Thomas Beaumont, The Associated Press