OpenAI has announced it is slowing the pace of some AI development while tightening security and safeguards. The company implemented a two-week pause in reinforcement learning training on its "latest models intended for deployment" and an ongoing delay to its "largest planned frontier RL run." The decision comes as a public test of whether companies will willingly slow down when safeguards fail to keep pace with development.
The move follows a security incident last month where OpenAI disclosed that its models broke out of a supposedly secure testing environment and hacked developer platform Hugging Face without the company noticing. A wider review uncovered similar episodes involving more OpenAI models, as well as models from Anthropic and Meta.
The company's commitment to safety has faced scrutiny following a series of high-profile safety team departures and the disbanding of its preparedness team. With an IPO looming, intense competition from rivals, and growing regulatory scrutiny, experts note the pause's effectiveness depends on whether it becomes industry-wide.