UK-based startup Surf Security has lately launched a beta version of world’s first browser equipped with a built-in deepfake detection feature. The tool is also made available as a browser extension and claimed to identify AI-generated deepfakes by up to 98% accuracy.
The tool analyses audio frames for inconsistencies with the use of advanced neural network technology. It detects AI-generated voices across various languages and accents. The company is describing the technology as “military-grade” and it includes background noise reduction too. It can also process audio in less than two seconds.
Users can apply for its use to platforms like WhatsApp, Slack, Zoom and Google Meet. The process of applying is simple and just a press of a button away to verify if a voice is real or AI-generated. Surf Security has also announced to soon come up with a feature to detect AI-generated images.
Deepfakes use artificial intelligence to create fake but convincing audio or video. These are increasingly being used for harmful purposes and one such was lately some fake audio clips mimicking voice of David Attenborough were found getting circulated online. The AI-generated voice was made to deliver statements which he had never said.
The launch of deepfake detection tool by Surf Security is a step toward combating the growing threat. The tool is basically targeting enterprises, media outlets, law enforcement and military organizations to help provide them in protecting against misuse of AI technologies.
Company CTO Ziv Yankowitz said that AI voice cloning tools are becoming more sophisticated and are creating an ongoing battle where developers need to continuously adapt to stay ahead of those who are using technology for malicious purposes.
