PHOENIX, United States — Edward R. Murrow once said, “To be persuasive, we must be believable. To be believable, we must be credible. To be credible, we must be truthful.” Nearly all media relied on that premise. Audiences trusted what they saw and heard because it came from real people speaking through trusted channels. That assumption no longer holds. Images, audio and video can now be manipulated to change what people appear to say or do, or completely synthesized without any humans ever involved. AI-generated media is now highly realistic and spreading faster than ever. A defining challenge today is distinguishing authentic digital media from content that only appears to be.
Authenticated visual provenance offers a path forward and is already deployed across a range of areas, including still images and video. A new approach based on audio extends this to verify the presence of a real human behind a microphone at the moment of capture. Together, they establish both where the media came from and that a real person spoke it.
Why does authenticity matter?
Authenticity has always mattered. The adage “seeing is believing” held for centuries. However, we are in a time when information velocity, combined with AI, is upending our senses and rapidly disrupting society, business and government.
For broadcasters and other high-trust media coming from government or corporate boardrooms, this shift is especially consequential. When the stakes are high, audiences prefer information, advice and recommendations from a real, accountable human. Historically, that trust was implicit: if you heard a voice or saw a picture, it meant a real person was behind the media.
By expanding what microphones capture, the industry can preserve evidence of how speech was produced, not just how it sounds, addressing the synthetic media problem at its core.
But not today. The World Economic Forum warns that fraud has become one of the most disruptive forces in the digital economy, eroding trust, distorting markets and directly impacting people’s lives. In addition to fraud, AI-based digital manipulation threatens the legal and institutional systems we depend on every day and significantly increases corporate liability. At a societal level, lost trust drives confusion and division. At the federal level, the United States has taken a firmly pro-AI stance, accelerating its development and innovation. This approach is likely to produce faster, more sophisticated and more adaptable AI systems that can be readily exploited for deception and fraud.
Just as automated bots forced the web to introduce CAPTCHA to authenticate users, the threat of synthetic media now demands new ways to establish both the provenance of content and the presence of a real human at the moment it is created.
Deciphering authenticity through provenance
Identifying AI-generated digital content is in high demand and will only increase as the proliferation of generative platforms continues. Detecting synthetic media after it has been created and released is highly challenging and, at best, probabilistic. Even under the best conditions, detection accuracy is limited — roughly 65% accuracy on untrained datasets, according to the research paper “Deepfake detection with and without content warnings” in the journal Royal Society Open Science — far less than usable at scale or in real-time. A more effective method is to verify authenticity at the time of creation. This approach, known as digital content provenance, is gaining ground across daily-use products such as Google Pixel phones, Ring cameras and software like Truepic Vision and Microsoft’s Azure cloud, as well as social media platforms like LinkedIn and YouTube.
These systems use interoperable cryptographic markings to ensure that authentic information from the moment of media creation is embedded in the file or fed into its system. This information, often referred to as “content credentials,” can help ensure authentication within the file, including the time, date, location and creation system of the media.
A critical component of provenance is device attestation. Pairing provenance with secure device verification establishes trust before content enters the digital ecosystem by ensuring authenticity is grounded in verifiable technical guarantees. With an attested image file, for example, recipients can be confident the image was not uploaded, injected, or altered, that the pixels have not changed, and that the time, date, location and orientation are authentic. Applied to audio, this would provide meaningful assurance, but a challenge remains with voice authenticity: How can a listener know a voice is from a live human and not synthetic or replayed?
Initiatives like iHeart’s “Guaranteed Human” reflect the industry’s desire to preserve trust. However, as networks scale across hundreds of stations, shows and contributors, it is difficult to reliably ensure content is human-generated.
In visual media, rebroadcast attacks can be detected by spotting a “screen of a screen.” Audio offers no such reliable cues. Once sound is captured, traditional systems lose visibility into how it was produced, making it hard to confirm that a real person was speaking, until now.
Broadcasting has always set the bar for trust in the media.
A new generation of hardware to address audio authenticity
To address this remaining gap in audio authenticity, verification must move upstream to the moment speech is created (similar to authentic video capture). Enhanced microphones offer our best opportunity to counter synthetic media by capturing speech the moment it becomes media. Human speech is not just sound. It is a physical process involving breath, vocal cord vibration and movements of the lips, jaw and tongue, yet traditional microphones record only the final acoustic signal.
For decades, this design was sufficient because voices only came from people. Generative AI has broken that assumption, leaving listeners unable to distinguish a live human voice from a cloned one. By expanding what microphones capture, the industry can preserve evidence of how speech was produced, not just how it sounds, addressing the synthetic media problem at its core.
OriginStory is developing capture hardware that enhances microphones to sense the physical biosignals of live human speech — signals that exist before sound becomes audio. The hardware is small and low-power and can be economically integrated into microphones, headsets, laptops and, eventually, phones.
As shown in the image above, an additional radar sensor next to the microphone captures uniquely human biosignals (lungs moving, vocal folds vibrating, etc.) in parallel with the audio. The audio is tightly coupled with these signals, creating a verifiable link between the voice and the human producing it, effectively binding the audio to a live person at the moment of recording.
Once audio is bound to a person at capture, new possibilities open across broadcast and distributed audio. Stations can authenticate on-air voices and endorsements, networks can enforce “human-only” content policies, and platforms can distinguish original human recordings from replay attacks or synthetic impersonations. For podcasts and on-demand audio, this same signal can support labeling, sorting, or prioritization based on verified human participation and, over time, inform sponsorship, pricing and licensing decisions. This human provenance can travel with the audio itself, embedded using watermarks or standards like C2PA, and surfaced through lightweight players.
Broadcasting has always set the bar for trust in the media. As synthetic media becomes indistinguishable from reality, the industry has a chance to lead again by pushing authenticity upstream to the moment of capture. Doing so will require new thinking about microphones, standards and verification, but it offers a durable way to protect voices, brands and audiences alike. Visual provenance has shown that this approach works at scale. Audio represents the next chapter, one with real promise to extend trust into the most human medium of all.
Mounir Ibrahim is the chief communications officer and head of public affairs and impact for Truepic. Visar Berisha is CEO and co-founder of OriginStory, which delivers proof-of-personhood authentication for digital communication.
This story originally appeared in the March/April 2026 edition of RedTech Magazine. You can read or download it for free here.
You can access all past RedTech publications, for free, here.
These stories might interest you
Five takeaways for radio’s next phase
Radio Luxembourg: How a small nation built a pan-European broadcasting powerhouse
