In the never-ending battle of figuring out what's real and what's created by artificial intelligence, NordVPN has a new browser tool that can help you separate human voices from fakes.
NordVPN's new tool listens to audio from your open browser window and alerts you if it detects a human or AI-generated voice. The feature is currently only available in NordVPN's Chrome browser extension, and you'll need to enable the tool to start detecting voices on websites in real time.
It isn't clear when the tool will roll out to other extensions, such as Firefox and Edge, or which plans include the AI Voice Detector. NordVPN didn't immediately respond to CNET's request for comment. In the meantime, here's what we know about the tool and how it works.
How NordVPN's AI Voice Detector works
NordVPN partnered with NordLabs' cybersecurity experts and tested both real and AI-generated audio. The tool uses an AI model to distinguish between the two types of voices, according to Thursday's announcement.
Here's how it works. You'll enable the feature in your Chrome extension, and the AI voice detector will run in the background on the tab you have open. You'll receive an update on the voice detector's results in the extension, and a pop-up that will display one of three colors from the sound:
- Green: It's a human voice.
- Amber: It could potentially be AI-generated.
- Red: It's an AI-generated voice.
NordVPN's announcement says that the detector cannot hear what's being said and that it only analyzes acoustic characteristics. The tool isn't capable of understanding, recording or interpreting conversations. You'll hear the sound as normal while NordVPN's detector analyzes it. If you stop the detection or close your active tab, NordVPN says all audio buffers will be discarded. Nothing will be stored.
NordVPN also called out the tool's security measures, including not collecting your identity, browsing history, cookies or account information. All of this makes me wonder about the false positive rate, given that NordVPN is analyzing sounds, not conversations.
Watch this: AI Is Indistinguishable From Reality. How Do We Spot Fake Videos?
How useful is an AI voice detector?
There are several AI detectors online, but some of them are trash. They can produce false positives (saying something is AI when it's not) or false negatives (saying something is human when it's AI) due to their inherent limitations.
In most cases, there are no foolproof watermarks or signatures to indicate that something is AI-generated. Some tools, like Copyleaks' AI detector, build more trust by showing their work, indicating which red flags they base their conclusions on.
NordVPN's tool might at least bring more skepticism and transparency for users, says CNET Senior Editor Moe Long.
"There's a lot of AI audio out there, and some of it may be increasingly difficult to identify by the human ear," Long says. "NordVPN's AI audio detector seems helpful because it can assist folks in determining whether certain audio is AI-generated or not, which could provide various benefits, from offering transparency about voiceovers in YouTube videos and on social media sites to avoiding scams."
Read more: AI Slop Is Destroying the Internet. These Are the People Fighting to Save It
AI-generated voices are increasingly used in predatory ways, like creating deepfakes of people for scams. Even a hint that not every voice you hear on the internet is genuine can help you spot fraud.
Long also likes that NordVPN's AI voice detector appears to offer safeguards that analyze audio while still maintaining privacy.
NordVPN scores highly for CNET, earning an 8.8 out of 10 on its CNET Score. The VPN company gets praises for its speed, helpful privacy features and large international server network. However, it falls short in renewal pricing and limited split-tunnel control for Android devices.
