DeepMark

DeepMark is a digital trust and media provenance company headquartered in Podgorica, Montenegro. It develops technology that embeds imperceptible, machine-readable identifiers into AI-generated audio, images, and video, and into live AI voice calls, so that receivers can later detect and verify whether content is synthetic or whether a voice agent is authorized. The company positions its technology as a way to support transparency obligations under the EU AI Act, particularly Article 50, and the multi-layer approach of the associated Code of Practice. DeepMark's origins lie in a university research lab in Montenegro, where two PhD researchers, Slavko Kovačević and Kosta Pavlović, spent years working on signal processing and audio watermarking. Their work grew into some of the first deep-learning approaches to audio watermarking, and by 2023 it had been cited as pioneering by researchers at Microsoft, Meta, Sony, and Adobe. That recognition helped drive the transition from academia to entrepreneurship, and DeepMark was officially launched with support from Montenegro's Innovation Fund and the Katapult Accelerator. The company offers two main pillars. The first is content marking and detection: AI-generated audio, image, and video can be marked with an imperceptible signal at generation time and detected on demand with clear results. The second is voice agent authentication: a verifiable, authorized-agent identity can be embedded into live AI speech so that any receiver can confirm a call came from an authorized agent, with the signal designed to survive VoIP, codecs, and the phone network. Both pillars follow the same four-step flow: register an identity or content provenance record and receive a credential; embed the signal at generation; detect or verify it later in a call, API, enterprise infrastructure, or network; and resolve the payload to a signed proof, such as "this is AI-generated and marked" or "this voice is an authorized agent." DeepMark serves generative AI providers, contact centers, banks and insurers, healthcare organizations, telecom platforms, and production companies. Its audio-watermarking technology is commercially licensed and used in production, and the company works with leading voice-AI and communications companies on production requirements. DeepMark has joined Y Combinator's Fall 2026 batch, which it highlights on its website as a signal of its next stage of growth.

DeepMark's journey began in a university lab in Montenegro, where two PhD researchers, Slavko Kovačević and Kosta Pavlović, spent years exploring signal processing and audio watermarking. Their initial curiosity centered on a technical question: how can information be hidden inside audio so seamlessly that humans cannot perceive it, yet machines can verify it? That research grew into some of the first deep-learning approaches to audio watermarking. By 2023, the work had been cited as pioneering by researchers at Microsoft, Meta, Sony, and Adobe. The recognition from major technology companies helped spark the leap from academia to entrepreneurship. DeepMark was officially launched with support from Montenegro's Innovation Fund and the Katapult Accelerator. The company is headquartered in Podgorica, Montenegro. DeepMark built its commercial offering around two pillars. The first is content marking and detection, which embeds an imperceptible, machine-readable signal into AI-generated audio, image, and video and allows detection on demand. The company states that this is built to support EU AI Act Article 50 transparency obligations and the Code of Practice's multi-layer approach. The second pillar is voice agent authentication, which embeds a verifiable, authorized-agent identity into live AI speech so that any receiver can confirm a call came from an authorized agent, with the signal designed to survive VoIP, codecs, and the phone network. Both pillars use the same four-step flow: register an identity, authorized voice agent, model, provider, or content provenance record and receive a credential; embed the signal at generation; detect or verify it later in a call, API, enterprise infrastructure, or network; and resolve the payload to a signed proof. The company says detection can work from as little as a few seconds of audio. DeepMark's audio-watermarking technology is commercially licensed and used in production. The company works with leading voice-AI and communications companies on production requirements and serves generative AI providers, contact centers, banks and insurers, healthcare organizations, telecom platforms, and production companies. DeepMark has joined Y Combinator's Fall 2026 batch, marking a new stage in its development.

• Research on deep-learning audio watermarking begins in a university lab in Montenegro

• Work cited as pioneering by researchers at Microsoft, Meta, Sony, and Adobe

• DeepMark officially launched with support from Montenegro's Innovation Fund and the Katapult Accelerator

• DeepMark joins Y Combinator's Fall 2026 batch

Digital trust and GenAI compliance infrastructure for verifiable media and authorized AI voice agents.

• Content marking & detection

  Marks AI-generated audio, image, and video with an imperceptible, machine-readable signal, then detects it on demand with clear results. Built to support EU AI Act Article 50 transparency obligations and the Code of Practice's multi-layer approach.

• Voice agent authentication

  Embeds a verifiable, authorized-agent identity into live AI speech so any receiver can confirm a call came from an authorized agent, surviving VoIP, codecs, and the phone network.

1. 1DeepMark | Verified Calls for the AI Voice Eradeepmark.me
2. 2About DeepMark | Invisible IDs for Digital Trust & GenAI Compliancedeepmark.me
3. 3DeepMarkwikidata.org