Why OpenAI’s Decision to Hold Back Voice Engine Is the Right Move

Last year, OpenAI introduced Voice Engine, an AI-powered tool capable of cloning a person’s voice with just 15 seconds of audio. The technology was groundbreaking—but also deeply concerning. A full year later, OpenAI has yet to release it widely, and honestly, I think that’s the right call.

AI-generated voices are already blurring the line between reality and deception, with scammers exploiting similar tools to impersonate real people. If an advanced system like Voice Engine were freely available, the consequences could be devastating.

The Growing Threat of AI Voice Cloning Scams

We’ve already seen how deepfake technology has been misused, but AI voice cloning is proving to be even more dangerous. Data from 2024 shows that AI voice cloning was the third fastest-growing scam of the year, and with good reason—it’s frighteningly easy to trick people.

Imagine receiving a call from what sounds exactly like your boss, a family member, or even a bank representative, asking you to transfer money or share sensitive information. How many people would be able to detect the scam before it’s too late? Probably very few.

Even today, with less advanced voice cloning tools, criminals have successfully used AI to:

• Bypass bank security checks that rely on voice authentication.
• Create fake emergency calls to manipulate victims into sending money.
• Impersonate politicians and celebrities to spread misinformation online.

Now, consider a tool like Voice Engine, which OpenAI itself admits is so realistic that it can replicate voices across languages and speaking styles. If it were released publicly, it could supercharge these scams to an unprecedented level.

Too Real to Detect: The Problem of Human Perception

One of the most alarming aspects of Voice Engine is that it sounds so real that the average person cannot distinguish it from a human voice.

Unlike AI-generated text, which can sometimes give itself away with odd phrasing or factual errors, AI-generated voices have no obvious tells. They don’t sound robotic. They don’t mispronounce words in unnatural ways. They don’t struggle with intonation.

This makes it nearly impossible for the average person to differentiate AI from reality—especially in high-pressure situations, like receiving a scam call from a voice that sounds exactly like a loved one.

Some of OpenAI’s proposed safeguards—like watermarking AI-generated voices and requiring explicit consent—are good ideas in theory. But how would they be enforced? Scammers don’t follow ethical guidelines. If Voice Engine were released widely, it’s inevitable that bad actors would find ways around these restrictions.

Final Thoughts: OpenAI Is Making the Right Call

I fully support OpenAI’s decision to delay (or potentially never release) Voice Engine. This isn’t about stifling innovation—it’s about recognizing the risks and acting responsibly.

AI-generated voices are already being misused, and a hyper-realistic tool like Voice Engine would only make things worse. Until there are effective regulations, detection methods, and enforcement mechanisms in place, releasing this technology would be reckless.

OpenAI has historically been accused of prioritizing rapid product releases over safety, but in this case, they’re taking a cautious approach—and that’s exactly what’s needed.

What do you think? Should AI voice cloning tools like Voice Engine ever be made widely available? Let’s discuss in the comments. 🚀

Citation: Wiggers, K. (2025, March 6). A year later, OpenAI still hasn’t released its Voice Cloning Tool. TechCrunch. https://techcrunch.com/2025/03/06/a-year-later-openai-still-hasnt-released-its-voice-cloning-tool/