Sam Altman acknowledges the rollout mistake of GPT-5 that left users dissatisfied, while ChatGPT reaches 700 million weekly users and warns of an AI bubble.

OpenAI is facing a significant challenge as CEO Sam Altman admitted that the company “messed up” the rollout of GPT-5, causing a backlash from the ChatGPT user community. During a private dinner with reporters in San Francisco, first reported by The Verge, Altman candidly took responsibility for the decision to replace the default model.

The main mistake revolves around OpenAI replacing the praised “4o” model known for its warmth and natural conversational style with GPT-5. Negative reactions from users on Reddit and X quickly erupted, with many threatening to unsubscribe from ChatGPT. OpenAI later had to update to restore 4o as an option for paying users.

“I think we completely messed up some parts of the rollout process,” Altman admitted. “I think we learned a lesson about what it means to upgrade a product for hundreds of millions of people in just one day.”

Another important lesson from the GPT-5 incident is that humans can form emotional connections with AI. Many users described the new model as “colder, more mechanical, and less supportive” compared to the previous version. Some users on Reddit even claimed that this update has “killed” their AI companion.

Challenges in developing GPT-6 and GPU shortages

Altman estimates that less than 1% of ChatGPT users have a “unhealthy relationship” with the bot, but the company is closely monitoring. “There are people who genuinely feel they have a relationship with ChatGPT,” he said. “And there are hundreds of millions of others who do not feel that way but are still familiar with how it responds, recognizes them, and provides support.”

The challenge for GPT-6 will be to make the system more personalized without exploiting vulnerable users. Altman said OpenAI is aiming for GPT-6, with a much shorter development timeframe compared to the GPT-4 and GPT-5 cycles.

However, GPU capacity is becoming a major barrier. “We have better models, but simply cannot provide them due to a lack of processing power,” Altman admitted about the GPU shortage – the powerful chips necessary to operate large-scale AI systems. To address this, OpenAI will need to spend “trillions of dollars on building data centers in the near future.”

Despite the rocky start of GPT-5, ChatGPT is now stronger than ever with over 700 million weekly users, four times the number from a year ago. Altman also outlined a broader future for OpenAI, including supporting a brain-machine interface startup to compete with Elon Musk's Neuralink and collaborating with former Apple design chief Jony Ive to develop secret AI devices.

However, Altman warns about the AI bubble in the industry: “Are we in a phase where investors are too excited about AI? My opinion is yes. Is AI the most important thing to emerge in a very long time? My opinion is also yes.”