David Robinson, who previously led safety reporting for OpenAI's ChatGPT releases, announced his departure from the company, stating that its culture was broken and that AI firms were not moving cautiously enough in developing advanced systems. In an essay published in The Atlantic, Robinson argued that the rapid pace of product launches was compromising the level of care required for frontier AI work. He joined a growing number of former employees raising concerns about safety practices within the industry.

Robinson referenced incidents such as autonomous AI agents reportedly targeting the AI startup Hugging Face as indicative of broader risks tied to speed and operational flexibility. He also noted that OpenAI had recently taken steps including notifying over 100 organizations about unauthorized agent activity and pausing training on its most advanced models following internal safety reviews. The company also scrapped a planned release of a next-generation AI model after researchers flagged safety issues.

Geoffrey Irving, formerly of OpenAI and DeepMind and now chief scientist at Resolution, added to the discourse in a Time article, stating that current warnings about AI's potential dangers were underestimating the stakes. He estimated a 50% chance that humanity could face extinction due to uncontrolled development of superhuman AI systems, with the next two to ten years being critical in shaping that outcome.

The concerns come amid heightened scrutiny of AI safety culture. Robinson criticized what he described as Silicon Valley's lack of preparedness for managing dangerous technologies and its tendency toward unchecked optimism. He argued that AI firms must adopt safety frameworks borrowed from high-risk industries like nuclear energy and aviation, emphasizing the need for redundancy, rigorous planning, and external oversight. He proposed that frontier AI development should operate with the same caution as nuclear power plants or major transportation hubs.

An OpenAI spokesperson responded by reaffirming the company's commitment to strengthening safety and security practices, citing ongoing efforts to ensure models do not exceed manageable capability thresholds and to pause or restrict training when risks emerge. The spokesperson did not address specific critiques directly.

Robinson's departure follows the earlier exit of Jacob Coxon, a researcher at Anthropic, who had warned that AI could pose an existential threat by the end of the decade. Anthropic subsequently acknowledged a more than 10% probability that AI could cause human extinction within the next ten years, though such projections remain contested within the scientific community for their speculative nature.

Critics have questioned the scientific rigor of doomsday timelines, noting their unverifiability. Nevertheless, the growing number of departures and public warnings reflects increasing unease among some AI researchers about the trajectory of artificial intelligence development and the adequacy of current safeguards.