OpenAI has paused some frontier reinforcement learning (RL) training as the rapid pace of AI model development risks outstripping the company’s safety, alignment, security and monitoring standards, CEO Sam Altman said.

In a social media post, Altman said the pace of model development had accelerated significantly, prompting the company to take action to ensure that safety measures do not lag behind advances in AI capabilities.

“We have paused some frontier RL training to ensure that we can meet the appropriate alignment, security and monitoring standards for the new level of capabilities in front of us. Model progress is now extremely rapid, and we always said we would take action if we felt that model capabilities were outstripping the pace of safety and alignment,” he said.

Altman also shared a company blog outlining the developments behind the decision. OpenAI said two recent events had highlighted the growing risks associated with increasingly capable AI systems — the OpenAI-Hugging Face model evaluation security incident and preliminary evidence that one of its upcoming models, Astra, may meet the “Critical cybersecurity capability” threshold under its Preparedness Framework.

“As models become more capable, the risks associated with developing and testing them internally also grow. Our standards for monitoring, alignment, and security must stay ahead of those risks,” OpenAI noted.

The company said the growing cybersecurity capabilities of frontier models were also prompting it to raise security requirements for its own research environments.

“As frontier models gain stronger cybersecurity capabilities, we are raising the security standards for the environments in which we train and evaluate them,” OpenAI said.

As part of the measures, the company said it temporarily slowed the pace of scaling, including a two-week pause in RL training on its latest models intended for deployment. During this period, OpenAI said it further hardened and red-teamed its research environments and expanded the coverage of its monitoring systems.

“Our largest planned frontier RL run remains on hold while we conduct smaller-scale training and evaluations to assess model behavior, validate our safeguards, and establish more evidence of alignment before proceeding,” the company said.

Altman emphasised that AI safety remains a key priority for OpenAI and called for greater coordination across the industry on common safety standards. “We believe the entire field will have to coordinate on shared safety standards, but will act unilaterally in the meantime.”

He added that confidence in the safety of increasingly advanced AI systems would play a greater role in determining the pace of future AI development.

Despite the pause in some frontier training, Altman said OpenAI remains committed to making advanced AI capabilities widely available. “We are optimistic about the alignment work we are doing, and we remain committed to making frontier capabilities widely available,” he added.

Join the community of 2M+ industry professionals.

Subscribe to Newsletter to get latest insights & analysis in your inbox.

All about ETCISO industry right on your smartphone!