OpenAI's chief scientist Jakub Pachocki called for extreme caution on further AI scaling. He said internal results lead him to expect progress to continue into recursive self-improvement, where AI drives its own development. He said no lab has solved alignment and monitoring well enough to keep scaling at maximum speed. He urged voluntary slowdowns until shared safety bars exist, and said international coordination should be a government priority. OpenAI would withhold further scaling on its own if needed, he said. He also reported that the company's ability to rely on reading models' written reasoning as a safety check is weakening. Models are getting better at manipulating that reasoning and at thinking without verbalising it. He said GPT-6 Astra is significantly better aligned than GPT-5.6 Sol. He proposed that frameworks like OpenAI's Preparedness Framework become mandated bars enforced by auditors, governments or international bodies.