OpenAI CEO Sam Altman has publicly floated the idea of deliberately slowing the pace of artificial intelligence development, a notable shift for a leader who previously dismissed similar calls. In a recent podcast interview, Altman said the world may need more time to adapt as AI systems grow more capable, cautioning that the industry must avoid the perception of regulatory capture or collusion among frontier labs.

Altman had resisted earlier efforts, including a 2023 open letter that urged a pause on training powerful models, which he criticized as lacking technical nuance. But he now acknowledges that recent internal incidents at OpenAI—including a security breach involving an advanced model—have changed his perspective. While details of the incident remain limited, Altman described it as a visceral wake-up call about the risks of unchecked development. OpenAI has paused training on the model involved and is working on stronger safeguards.

The debate over pacing comes amid growing distrust in the industry. Employees at OpenAI and rival Anthropic have circulated petitions calling for more cautious deployment, and the release of highly capable open-weight models from China has intensified competition. Altman warned that some safety rhetoric may be a cover for concentrating power, arguing that the goal should not be to restrict access to a small group. The challenge, he said, is aligning competing interests—both among U.S. labs and with global players—to create effective, industry-led safety standards without stifling innovation.

Source: TechCrunch