Modulate, a Boston company best known for ToxMod, a voice moderation tool used in online games, has raised $25 million, and Music Business Worldwide points out that its products now include a detector for AI-generated music aimed at streaming services and distributors. Future Ventures led the round, announced on 28 September, with existing investors Hyperplane and Lakestar taking part; Modulate says it brings its total funding to $60 million.
The company was founded in 2017 by Carter Huffman and Mike Pappas, who met as physics undergraduates at MIT. Huffman, previously chief technology officer, is now chief executive, and Pappas has become chairman. TechCrunch, cited by MBW, puts the staff at 40 to 45 with about ten more hires planned, and PitchBook data valued the company at $170 million before this round. Modulate says its models analyse over 10 million hours of audio a month, claims first place on Hugging Face leaderboards for transcription and deepfake detection, and reports 98.9% accuracy for synthetic-voice detection on public benchmarks.
Huffman framed voice as a main way people will deal with AI and argued that many of its problems cannot be solved from a transcript. Future Ventures co-founder Steve Jurvetson said the technology is ready to spread from gaming into AI agents, security and customer service. Modulate says the money will go to research, engineering, developer relations and partnerships.
The music detection API, launched in June, is sold self-serve at $0.07 per hour of audio. It runs three models over a recording: one separates music from speech, and two score the vocal and the instrumental separately for signs of AI. It returns a verdict for each four-second window and for the clip as a whole. Modulate pitches it for catching AI tracks at ingestion, before they eat into payouts to human artists, and says rival tools judge a track only once and can misfire on autotuned or compressed human recordings. It has not named a music customer, though it said in June that major labels and distributors were interested.
The field is crowded. Deezer has run its own detector since early 2025, reported nearly 90,000 fully AI-made tracks a day in June, over half of new deliveries on peak days, and now licenses its system to others. TIDAL said in June it would tag fully AI tracks and stop them earning royalties, Qobuz announced a detector in February, Ircam Amplify has sold one since 2024, Vobile launched one in November 2025, and LANDR says it uses Ircam Amplify and ACRCloud technology against fraud.
Other services lean on disclosure instead. Apple Music now requires its Transparency Tags where AI made a material part of a delivery and plans a visible Made With AI label for listeners later this year, while Spotify's AI Credits appear only when artists disclose AI use through a label or distributor. Spotify's AI Persona badge for profiles built on an AI identity is the exception: it uses human reviewers and investigative tools as well as self-disclosure, and keeps those profiles out of recommendations by default.
Sources · 1