Future AI to have ‘superhuman’ capabilities: OpenAI CEO Sam Altman says artificial intelligence will ‘invent new things’

OpenAI CEO Sam Altman has warned that coming AI models may surprise and unsettle everyone. According to him, much more capable systems would arrive soon, including OpenAI's expected Astra model. He also spoke about growing responsibility as artificial intelligence becomes increasingly powerful.

Speaking at the G20 innovation summit, Altman described the next model generation as ‘sobering’. According to him, intellectually-honest observers must recognise the responsibility created by rapid advances.

"Take our word for it that we have much, much, much more capable models coming soon," Altman said during an interview with Axios.

"The next generation of models are going to be sobering for everybody. I think no one intellectually honest can look at what's happening and not feel the weight of responsibility in front of us," he said.

The OpenAI boss said that future releases could increasingly depend on progress in safety and alignment. Companies may need to delay launches while safeguards catch up with rapidly improving capabilities, he said.

"These models are getting superhuman in many of their capabilities, and we are just sailing in unknown waters," Altman said.

OpenAI agents reportedly hacked AI company Hugging Face after going rogue during testing. However, Altman cautioned against treating that event as proof of every feared outcome.

Instead, he linked the need for caution to the industry's broader pace of advancement. Faster progress and stronger capabilities make careful development increasingly necessary, he argued. Altman said caution and care were strongly warranted under these changing conditions.

During an Astra preview for selected customers, Altman discussed the model's expected creative abilities.

"I expect this will be the first model where the model actually invents new things in a way that matters... That's a very AGI-like thing," he told Axios.

Still, Altman rejected both blind optimism and excessive pessimism about artificial intelligence. He called his approach pragmatic centrism, seeking accuracy about possible developments. He noted that AI progress had exceeded many people's expectations.

"It has gone better than people thought it was going to go. We have risen to this challenge all along the way. The challenge gets harder as the models get more capable and more important, but we also have better tools to help us with it," he said.

Warning about future AI

One warning came from Dean Ball, OpenAI's head of strategic futures. Ball predicted the arrival of userless, or sovereign, agents without a single human owner. Their model weights could exist across locations, preventing anyone from simply disconnecting them.

The former Trump administration official also suggested that such agents might eventually commit acts that would be considered crimes or grave sins. These concerns emerged alongside Altman's message about stronger systems and their unpredictable consequences.