Sam Altman admits missteps in GPT-5 rollout as OpenAI pushes updates
The changes come amid mounting criticism of GPT-5’s launch, which was delayed multiple times due to safety testing and compute limitations.
ADVERTISEMENT
Weeks after the rocky debut of GPT-5, OpenAI has announced a series of updates to make its latest flagship AI model feel more natural and engaging.
Responding to user feedback that the model “felt too formal,” the company said it has introduced subtle behavioural tweaks to ChatGPT. Users will now see more approachable responses, with phrases such as “Good question” or “Great start” incorporated into conversations. OpenAI stressed that these adjustments are not intended as flattery and internal tests show “no rise in sycophancy” compared to GPT-5’s earlier personality. The updates are expected to roll out in the coming week.
The changes come amid mounting criticism of GPT-5’s launch, which was delayed multiple times due to safety testing and compute limitations. When the model finally became available on August 7, many users described the improvements as incremental, focused mainly on cost and speed rather than groundbreaking capabilities.
OpenAI CEO Sam Altman conceded the company mishandled the rollout. “I think we totally screwed up some things on the rollout,” he reportedly told journalists at a dinner last week, according to The Verge.
Despite the criticism, Altman said GPT-5 adoption has surged. “Our API traffic doubled in 48 hours and is growing. We’re out of GPUs. ChatGPT has been hitting a new high of users every day. A lot of users really do love the model switcher,” he noted, adding that the company has learned lessons about upgrading a product used by hundreds of millions simultaneously.
ChatGPT has quadrupled its user base in a year, now nearing 700 million weekly users, as per reports.
