Deepfakes Decouple Authenticity From Trust
Platforms Strip C2PA Manifests
Platforms like Meta (Instagram, Facebook, Threads), X (formerly Twitter), TikTok, WhatsApp, Reddit, and Discord routinely remove C2PA manifests during standard upload processing through re-encoding pipelines, file size optimization, and privacy protection workflows. Diva-portal observed, however, that detection alone cannot be the sole foundation of trust because deepfake generation scales faster than detection. Mohanad Alamin and his co-authors pointed out that misleading or out-of-context content can still be cryptographically signed and distributed, undermining confidence in editorial judgment. The CJID explained that these standards struggle to overcome systemic skepticism because their enforcement is often voluntary, inconspicuous, and easily stripped during content distribution.
Platforms Strip Authenticity Data
The CJID documented that C2PA content credentials and other embedded metadata are routinely stripped by major social media and messaging platforms, severely compromising the effectiveness of hardware-backed signing mechanisms. Most major social media and messaging platforms process uploaded media through compression, resizing, and format conversion pipelines, which routinely strip embedded metadata, including C2PA content credentials, a fact corroborated by a Florida State University repository, Mohanad Alamin and his co-authors, Diva-portal, Taylor & Francis Online, AFIP, and MetaClean.
The Liar's Dividend Accelerates Distrust
Mohanad Alamin and his co-authors, along with Wikipedia, observed that the "liar's dividend", where public figures strategically dismiss authentic but damaging recordings as AI-generated, further exploits this uncertainty, eroding accountability and institutional trust by creating a continuous state of doubt. Mohanad Alamin and his co-authors elucidated that deepfakes erode institutional trust by imposing a "skepticism tax," which creates systemic doubt about the authenticity of all media, extending beyond specific instances of synthetic disinformation. The CJID reported that this phenomenon means even rigorously verified journalistic content is viewed with suspicion because deepfakes decouple visual evidence from truth. Mohanad Alamin and his co-authors determined that this systemic doubt disproportionately affects politically moderate individuals and shifts audience reliance away from established news organizations toward closed, peer-based networks.
Human Deepfake Detection Fails at 55%
Nature, Diplomacy, and The CJID all indicate that human detection accuracy for sophisticated deepfakes remains near chance, averaging between 55% and 57% even among trained observers. Expecting the general public to reliably discern truth places a heavy cognitive burden on those less equipped to do so. Current governance strategies that heavily rely on public media literacy disproportionately burden demographics with lower digital literacy and higher visual trust. Mohanad Alamin and his co-authors documented that older adults, for instance, share nearly seven times more misinformation than younger cohorts due to lower digital literacy and greater reliance on visual evidence. An industry survey by MetaClean reports that older adults are particularly vulnerable due to lower digital literacy.
Trust Decouples from Truth
The persistent erosion of trust, fueled by platform actions and cognitive biases, means cryptographic provenance alone cannot bridge the chasm between technical authenticity and public confidence. This dynamic implies a growing societal divide in access to reliable information, potentially leading to further polarization and reduced civic engagement among vulnerable populations.
Comments ()