Sam Altman during the OpenAI DevDay 2026 keynote in San Francisco, 29 September 2026 Credit: Screenshot: OpenAI DevDay livestream OpenAI now has an AI research intern, chief executive Sam Altman said on Tuesday. He made the claim in his DevDay 2026 keynote in San Francisco. He told 2,500 developers that the company predicted a year ago it would reach that goal within 12 months. “A few weeks ago we reached that goal. We now have an AI research intern,” Altman said.
Tejal Patwardhan, a research lead for post-training at OpenAI, followed him on stage. She said its models now complete more than a third of day-long research tasks with no human help. She said the models have already helped solve more than 100 maths problems that had been open for decades. Dots and the keynote launches Altman opened the keynote with always-on dots agents, which run on their own cloud computers.
Users can call them, and texting is coming soon. “I have been waiting for this product for years,” Altman said. “It’s the way I’ve always wanted to work with AI.” He also launched a cheaper GPT-6.1 Sol model, a $500 Pro plan and a faster Ultrafast tier. Altman said its API keeps more than 99% reliability. It has also cut the time to the first token of a response by 45%. The full list of launches is in our DevDay round-up.
He closed the keynote on what he wants AI to become. “If we get this right, [the future] can be more like a Renaissance than an industrial revolution,” Altman said. Safety, before the keynote Before going on stage, Altman spoke to Kate Rooney on CNBC’s Halftime Report, according to a transcript published by CNBC. “Alignment, safety, monitoring, security have to stay way ahead of capabilities,” Altman told CNBC. OpenAI decided this week to shelve its Astra update, GPT-6.1 Astra. Altman said the model was a little worse on a few of the company’s evaluations.
He put the decision in the “abundance of caution category”. Altman told CNBC he was not aware of any incident as serious as the Hugging Face breach. He called Nvidia’s new agent safety platform a good thing, and said OpenAI is doing similar work. “I don’t think it’s a full solution,” he told CNBC. He said AI safety is also a scientific problem, not only an engineering one.
IPO and hardware Altman told CNBC he does not “have a particular timeline in mind” for an initial public offering. He said OpenAI could go public once it understands how to handle the next level of AI safely. On hardware, he said OpenAI will have “something that’s worth waiting for”, but not on Tuesday.
