OpenAI delays ChatGPT's Advanced Voice Mode

That hyper-realistic voice demo OpenAI showed off at its Spring Update needs a little more time in the oven.

OpenAI announced that it is delaying the alpha rollout of ChatGPT's advanced Voice Mode. The company originally planned to launch it to a small group of ChatGPT Plus subscribers in late June, but said it needs one more month to reach its launch bar.

Safety and servers need work

According to OpenAI, the extra time is going toward beefing up safety and scale. The team is currently improving the model's ability to detect and refuse certain content, smoothing out the user experience, and prepping infrastructure to handle millions of real-time requests without lag.

The iterative rollout will now kick off with a small alpha group to gather feedback, with general availability for all Plus users planned for the fall. Meanwhile, the video and screen-sharing features previewed during the same event are still in the works and will arrive on a separate timeline.

Why it matters

Live demos on stage are one thing; rolling out real-time conversational voice to millions of unpredictable users is an entirely different beast. We think OpenAI made the right call by pumping the brakes here.

Voice is inherently much riskier than text. When an AI can convey emotions and respond with zero latency, the surface area for emotional manipulation, safety bypasses, and impersonation grows exponentially. Plus subscribers will have to wait until fall, but it is far better to get a rock-solid feature late than an uncontrollable toy early.