Spotify Acquires Multilingual Harmful Content Detector Kinzen
Spotify has purchased Kinzen, a startup that uses machine learning to identify and deal with problematic audio content in multiple languages, for an unrevealed sum. The audio platform has worked with Kinzen since 2020 to detect and address hate speech and harmful content in the podcasts and shows it hosts using both AI models and human expertise to “more effectively deliver a safe, enjoyable experience on our platform around the world,” according to Spotify.
Spotify streams audio around the world in hundreds of languages and dialects, an impossible amount of audio to comb through manually. Kinzen uses AI to transcribe and analyze audio content in many languages for “dangerous misinformation” that it then flags for review. Kinzen employs and retains a global network of experts and native speakers of those languages to determine if there really is a problem and to refine its models as they parse the shows.
“We’ve long had an impactful and collaborative partnership with Kinzen and its exceptional team. Now, working together as one, we’ll be able to even further improve our ability to detect and address harmful content, and importantly, in a way that better considers local context,” Spotify global head of public affairs Dustee Jenkins said. “This investment expands Spotify’s approach to platform safety, and underscores how seriously we take our commitment to creating a safe and enjoyable experience for creators and users.”
Content moderation with AI is on the rise as any meaningful human-only efforts become impossible. Kinzen’s efforts are reminiscent of startups like Modulate and its ToxMod AI platform, which recently raised $30 million. ToxMod is embedded in game platforms to flag violent or otherwise offensive speech in real-time video game voice chat. Similarly, AI content moderator Spectrum Labs raised $32 million in January for its text and voice-based AI moderator that it claims can recognize and categorize more than 40 kinds of toxic comments. Spotify went for an acquisition, but Microsoft has filed a patent for an AI to measure how people are feeling based on their voice and other signals to moderate voice chat on its game servers, and Intel has developed a voice AI tool called Bleep on Intel PC chips to spot and censor offensive language uttered by a player before anyone else hears it.
Ultimately, Spotify hopes acquiring Kinzen will give it a better foundation for quickly and efficiently dealing with issues of abusive or deliberately misleading information before they become a moderation and public relations problem. Though it doesn’t seem to have too much of a negative financial impact, Spotify is likely not keen to repeat the headlines and backlash from famous artists it faced after Joe Rogan repeatedly shared vaccine misinformation on his Spotify-exclusive podcast.
“The combination of tools and expert insights is Kinzen’s unique strength that we see as essential to identifying emerging abuse trends in markets and moderating potentially dangerous content at scale,” Spotify head of trust and safety Sarah Hoyle said. “This expansion of our team, combined with the launch of our Safety Advisory Council, demonstrates the proactive approach we’re taking in this important space.”