The phone rings. A customer service voice answers, warm and efficient, and somewhere in the middle of the conversation you realize you are not talking to a person. ElevenLabs is likely on the other end, and its CEO recently told us that businesses should probably disclose that fact, at least until interacting with a machine becomes the default expectation. That last clause is doing a lot of work, and it deserves a closer look. We are moving toward a world where the artificiality of the voice is not a bug to be hidden but a feature to be managed, and the question of when to tell you is really a question about how much trust we are willing to outsource.
The CEO's position is pragmatic, and that is what makes it interesting. He is not arguing for transparency as a moral imperative; he is arguing for it as a practical bridge to a future where the disclosure becomes unnecessary because everyone will assume a machine on the line. That is a subtle but important distinction. It suggests that the current expectation of human contact is a temporary condition, one that will fade as the technology becomes more capable and more common. For our readers who are building or deploying AI systems, this is not a distant philosophical debate. It is a design decision you are making right now, whether you have named it or not. The same logic applies to other domains we have covered, like the practical challenges in Exploring Real-World Computer Vision, where the gap between a promising demo and a reliable deployment is filled with unglamorous tradeoffs. Disclosure is one of those tradeoffs, and it is not a technical one.
Here is our honest take: if you are a business using AI voices, treat disclosure not as a legal hurdle but as a feature of your product. The moment a customer feels deceived, you have lost more than you gained by hiding the machine. The CEO is right that the expectation will shift, but that shift is not inevitable. It is shaped by the choices companies make today. If you are transparent now, you build a reputation for honesty that will carry over when AI voices are the norm. If you hide it, you are training your customers to be suspicious of every interaction, including the ones that are genuinely human. We would tell a reader who asked us directly: do not wait for a regulation or a competitor to force your hand. Build the disclosure into the experience now, not as an apology but as a matter of course.
The deeper issue is that we are not just building tools; we are reshaping a social contract. The Forrester Function might help you optimize a model, but no amount of tuning tells you when to say "I am a machine." That is a human judgment call, and the CEO's comment suggests the industry is just beginning to figure that out. The specific detail to watch is the phrase "until everyone expects it anyway." That is a moving target, and it will not move on its own. It will move because companies like ElevenLabs decide to be upfront, or it will move because the rest of us get tired of being fooled. The question is not whether the expectation will change. It is whether you want to be part of the reason it changes for the better, or the reason it changes for the worse. The voice on the phone is getting harder to place, but the responsibility for what it says is still entirely human.