OpenAI says AI can make itself safer but urges global safety standards first

OpenAI stated it aims to build an automated AI researcher to address the alignment problem, asserting this could eventually make AI safer, but warned such development must not proceed until it can be done safely. The company cautioned that recursive self-improvement (RSI), where AI trains itself without human oversight, risks humans losing practical control, potentially making AI more dangerous and less aligned. It urged for new national and international safety standards, plus better measurement and incident reporting protocols. OpenAI specifically requested the US government to lead these safety efforts due to its frontier AI industry and global position, despite President Donald Trump having rejected similar calls from companies like Anthropic, citing concerns about ceding leadership to China.