GPT-5 can indeed produce impressive results in benchmarks compared to competing AI models, but a lot of users have claimed that it's hardly an improvement over its predecessor, GPT-4o. Users have been flagging issues like bugs, glitches, and unresponsiveness since the rollout of GPT-5.

Beyond the immediate user experience, Altman’s concerns concern the ethics of AI as a therapeutic tool. He is worried about AI unknowingly pushing a user away from their long-term well-being.