OpenAI CEO Sam Altman has acknowledged growing anxiety over artificial intelligence as increasingly capable models reach the public. He said even people working closely with AI feel both excitement and concern about its rapid progress. His comments come as OpenAI prepares its next model while facing questions over AI safety, data centers and public trust.
Altman said OpenAI has experienced an uneasy balance between enthusiasm about new AI capabilities and concern about where the technology could lead. He also acknowledged that people outside the industry may feel even more uncertain.
“We have been living with the tension between being excited and anxious about progress for some time,” Altman said. He added, “We know it is much more discordant for other people.”
His strongest warning focused on the uncertainty surrounding advanced AI. “AI is getting extremely capable; no one fully understands the consequences of this,” Altman said. He argued that managing the transition toward widely available and powerful AI should become a global priority.
However, Altman said slowing technological progress alone would not solve the problem. He believes society needs to understand how AI systems work in real-world settings while developers improve safeguards.
OpenAI has spent much of the summer working on safety measures as its models become more capable. Altman said the company wants safeguards and AI capabilities to advance together.
The company has also slowed some development work when additional safety testing is required. “We are clearly in a phase of development where we believe caution is warranted,” Altman said.
According to Altman, OpenAI has been ‘pacing’ its progress so new systems can meet safety standards tied to higher capability levels. He said safety and alignment work has also affected the development timeline for models that will follow its upcoming release.
Altman described alignment as a central part of managing more powerful AI. In this context, alignment refers to efforts to make AI systems behave according to intended goals and safety requirements.
Altman also discussed Astra, an OpenAI model that he said has already completed training. He described it as a step forward in both capability and alignment.
“There is an obvious tension here: on one hand, Astra is very good and we are excited to see what people will build with it,” Altman said. “On the other hand,” he said, OpenAI believes greater caution is now required.
OpenAI plans to launch its next model soon, according to Altman. However, he said the company has slowed work on models beyond Astra when needed to allow more time for safety and alignment research.
Altman’s comments also come as AI companies face resistance over the infrastructure needed to train and operate large models. Data centers require large amounts of electricity and, in some cases, water for cooling. Communities have raised concerns about energy costs, resource use and new construction.
“Clearly, people hate data centers — right now, at least,” Altman told Time in a recent interview. He also acknowledged broader public sentiment, saying, “People are pretty negative on AI.”
Despite that resistance, OpenAI continues to invest in computing infrastructure. Altman has argued that companies developing advanced AI must address safety while society adapts to the technology.
He said an ‘iterative loop’ in which AI and society evolve together offers a path for managing the transition. Still, his comments make clear that uncertainty remains. Altman concluded, saying, “Getting AI safety right is more important than any company’s momentum.”
ALSO READ: Apple, OpenAI Clash Over Trade Secrets as New Evidence Emerges