OpenAI has decided to postpone the launch of voice and emotion-reading features for its ChatGPT chatbot due to the need for safety testing. The decision comes after a demo last month sparked interest from users but also drew a threat of a lawsuit from Scarlett Johansson over alleged voice copying. Originally set for a June release to paying subscribers, the new features will now be available in the fall, with exact timelines depending on safety and reliability standards. This delay highlights the challenges tech companies face as they strive to advance generative AI technology. OpenAI's cautious approach reflects a growing awareness of the potential pitfalls of AI, as seen in recent incidents involving Google and Microsoft. The company aims to fine-tune the chatbot to prevent issues like spreading false information or bias. By prioritizing safety and reliability, OpenAI hopes to deliver a successful and ethically sound product to its users.