VCs like Clarity’s approach to deepfake detection, but there are alternatives

Ideas for detecting deepfakes are popping up among startups and attracting significant venture funding.
Just a year ago, most news about spotting AI fraud came mostly from university research labs. Now, everyone has numbers showing how common deepfake fraud attempts have become.
And despite how rare it is to meet someone who’s witnessed an attack, money is flowing to software firms to defang fraudulent video, images and audio, especially by detecting them.
Clarity, created in 2022 to disrupt the criminal world’s deepfaking efforts, has closed a $16 million seed round. That’s not huge, but there are many, many biometric entrepreneurs who’d love to close a round like that right now.
The leaders of the round were Bessemer Venture Partners and Walden Catalyst Ventures. Participants included Flying Fish Partners, Ascend Ventures and Secret Chord Ventures.
Technology investment publisher TechCrunch has posted an interview with Michael Matias, co-founder and CEO of Clarity.
The company has a story to tell – clearly – but maybe not the kind some in biometrics would think of.
Matias says Clarity is run like a cybersecurity company. Rather than viewing the deepfake world’s dynamic as waves, he sees viruses constantly evolving. To battle them takes the same kind of continuous adaptability.
So far, at least five companies with deep pockets like the approach.
Another point of view could collectively be described at watermarking, making it more obvious that audio, images and video online is verified as genuine.
Mashable, a technology news and culture publication, has taken a look at Hugging Face, a developer of machine learning tools and proponent of ethical AI development.
Hugging Face is branching into watermarking for deepfakes. It’s a good idea, but it’s common to run across, say, a snippet of war footage, which is genuine and could be watermarked, but out of context. It was recorded in another conflict entirely.
Information consumers globally rarely investigate further, potentially making watermarks less effective as a defense.
Verification platform firm Sumsub has its own idea. In fact, the company is claiming an industry first.
Its detector is integrated into the company’s video identification and biometric liveness software. (Sumsub has its own statistics with alarming figures on the prevalence of deepfakes.)
The software works in real-time during interviews with people. The company says it was performing identity checks for a client that needed video verification of its end users.
According to Sumsub, it was apparent that the interview process was “susceptible” to attack. From there, executives decided to embed deepfake detection into the video verification process.
Detection is at the stage where most product ideas seem at least plausible, but one of the better ways to find substance is to follow the money.
Article Topics
biometric liveness detection | Clarity | deepfake detection | funding | Hugging Face | identity verification | Sumsub
Comments