Why Text Was the Missing Piece in Content Authenticity

By: Erik Svilich, Founder & CEO | Encypher | C2PA Text Co-Chair

*On January 8, 2026, the C2PA 2.3 specification officially published—and with it, Section A.7: "Embedding Manifests into Unstructured Text." 󠇟󠇠󠇡󠇢󠇅󠇯󠄈︉󠇇󠆬󠄹󠄹󠆪󠅊󠇫󠅹󠄽󠄠󠅎󠆉󠇮󠆸󠄣󠄿󠅦󠄚󠆃󠄩󠄁︂󠅦󠄼󠆪󠅙︋󠅷󠆱󠆙󠄩󠅵󠄨󠅪󠇢󠇆Text finally has provenance. 󠇟󠇠󠇡󠇢󠄮󠇓󠄈󠇬󠆶󠇯󠄳󠅗󠆂󠇉󠆣󠄞󠇪󠅇︋󠆛󠇐󠄒󠆄󠆃󠅃󠄣󠇛󠇏󠄠󠅿󠆺󠄤󠄜󠄅󠄍󠄘󠆳󠆭󠅛󠆔󠇭󠆀󠇤︋Here's why this matters. *

󠇟󠇠󠇡󠇢󠆄󠆵󠆺󠅇󠅪󠇦󠄲󠆹󠅱󠅪󠇎󠆾󠇆󠆀󠆂󠄍󠅸󠄲󠆳󠄊󠄦󠅒󠇟󠅕󠅞󠄳️󠇅󠆑󠇧󠆖󠇂󠄑󠄄󠆍󠆆󠄅︍󠇢󠄕For years, the content authenticity community focused on images and video. 󠇟󠇠󠇡󠇢󠅰󠄄󠄡󠇙󠄢︊󠄳󠆔󠆤󠄃︉󠆐󠇝󠆎󠅕󠅈󠅾󠆜󠇂󠄇󠇙󠄌󠆰󠇞󠄲︃︇󠆥󠅨󠅟󠇮󠅆󠄒󠄬󠇉󠆮󠅂󠇠󠇏󠆗EXIF data, C2PA manifests, watermarking—all designed for visual media. 󠇟󠇠󠇡󠇢󠆒󠇆󠅘󠇖󠆅󠄺󠄸󠆵󠆡󠆗︊󠄩󠆸󠄀︎󠆻󠅙󠇍󠅊󠆾󠇓󠆇︂󠄯󠄛󠅗󠇯󠄆󠄿󠄼󠆇󠆬󠅫󠇣󠆙󠆦󠅃󠄚󠄏󠅵Meanwhile, text—the format that makes up the vast majority of AI-generated content—had no provenance solution at all. 󠇟󠇠󠇡󠇢󠇨󠅑󠆋󠆿󠄇󠇎󠄿󠅄󠆘󠅜󠇕󠅛󠇊󠆚󠄴󠇪󠅘󠅊󠅎󠆗󠄕󠄺󠇒󠆥󠆹󠆊︂󠅑󠄲󠄌󠇁︉󠄦󠄚󠇫󠅟󠆛󠄻󠇞󠅈This wasn't an oversight. 󠇟󠇠󠇡󠇢󠅅󠅬󠆼󠅩󠇚󠇐󠄸󠇂󠆙󠆋󠆀󠄈󠇨󠄧󠇛󠄷︅󠆍󠇙︈󠇖󠇠󠇕󠆎󠇆󠄻󠄿󠅸󠇊󠆦󠆅󠅊󠆜󠅎󠇋󠅃︈󠆰󠆁󠄋It was a genuinely hard problem. 󠇟󠇠󠇡󠇢︈󠆢󠄗󠇧󠄬󠅗󠄸󠇙󠆆󠅱󠅤󠄵󠆴󠆁󠇟󠅻󠅎󠅗󠇨󠇝︃︄󠅣󠅶󠇘︉󠄜󠅻󠆍󠆰󠆏󠇇󠄐󠅧󠅻󠇕󠄠󠇞️󠇒And solving it required rethinking how we embed proof of origin into the most fundamental form of digital communication. 󠇟󠇠󠇡󠇢︁󠇊󠅙󠆪󠆑󠆽󠄶󠆲󠅷󠆫󠅎󠄱󠄥󠅟󠄔󠄿󠄖󠅮󠄰󠄟󠅽󠄅󠇩󠆄󠇝󠅕󠄢󠆦󠄿󠆈󠅾󠇫󠆼󠄠󠄦󠇚󠆠󠅻󠅐󠇕## The Content Authenticity Landscape Before Text When the Coalition for Content Provenance and Authenticity (C2PA) launched in 2021, it addressed a clear need: proving where images and videos came from in an era of deepfakes and AI manipulation. 󠇟󠇠󠇡󠇢󠄻󠇌󠆂󠇫󠅟󠆕󠄰󠄙󠆅󠄀󠆰︈󠅎󠆤󠄇󠄵󠄧󠅿󠅞󠆨󠅓󠆕󠆈󠄓󠄤󠄠󠄖󠅷󠆩󠅼󠄏󠅣󠅏󠄡󠇙󠅈󠆎󠆼󠇖󠄜### What Existed

Images: EXIF metadata had been around since the 1990s. 󠇟󠇠󠇡󠇢󠅕󠄛󠅣󠆓󠄵󠅭󠄾󠄰󠅼󠅙󠆂󠅐󠄬󠅌󠇊󠇉󠅚󠆅󠆣󠅪󠄺󠄀︆󠆜󠄮󠅠󠄌󠆟󠇞󠆉󠇮󠄠󠄟︉󠅑󠄓󠆘󠅷󠆠󠆕C2PA built on this foundation with cryptographic manifests that could prove origin and track edits. 󠇟󠇠󠇡󠇢󠆛󠅁󠆀󠅄󠄸󠄓󠄵󠅣󠅿󠇙󠇩󠆧󠅒󠄫︇󠅝󠆩󠄣󠇌󠄓󠆻󠆁󠅾󠅌󠆠︌󠇀󠄀󠅋󠆿󠅰󠄬󠆎󠇫󠄍󠅗󠄗󠅪󠅩󠇚Video: 󠇟󠇠󠇡󠇢󠅰󠆋󠄆󠇛︊󠅀󠄻󠆣󠆓󠆷󠇘󠆢󠅞󠅘󠅇󠆙󠇬󠇔󠆝󠄾󠇀󠄭󠅥󠅬󠆈󠄤󠆔󠆗󠇉󠇯󠄛󠄋󠅴󠅍︂󠅤󠆪󠆏󠆻󠅕Container formats like MP4 and MOV had metadata structures. 󠇟󠇠󠇡󠇢󠄸󠆴󠄚󠅽󠄫󠄇󠄸󠆺󠆮󠆙󠇫󠄿󠄎󠇤󠅥󠇝󠄕󠆫󠄭󠅺󠄴󠇒󠆋󠇠󠄈󠆏󠆳󠅵󠆘󠅧󠆟󠇃󠇩󠄏󠅍󠅩󠇝󠅛󠄸󠇎C2PA extended these with signed manifests. 󠇟󠇠󠇡󠇢︆󠄎︂󠇀󠄏󠇂󠄼󠇩󠆚󠄉󠆗󠇖󠆠󠆯󠅛󠄏󠆗󠆍󠆀󠆾󠅍󠄋󠆥󠇦󠆇󠆑󠆷󠆚󠄞󠇦󠄃󠅧󠆉󠅭󠄟󠄘󠇂󠄗󠆺󠅑Audio: Similar to video, audio containers could hold provenance data. 󠇟󠇠󠇡󠇢󠆏󠆽󠅜󠆈󠇧󠆬󠄺󠆬󠆍󠆝󠄩󠆧󠆍󠇞󠅚󠆮󠆉󠄟󠅁󠆽󠅻󠇂︆󠇀󠅡󠇨󠇠󠇥󠄼󠇪󠄏󠆺󠇭󠆕󠇙󠄈󠅸󠇈󠆣︂PDF: 󠇟󠇠󠇡󠇢󠄪󠇋︆󠅩󠆙󠇓󠄻󠅉󠆙󠅽󠆾󠄨󠅃󠇖󠇜󠇏󠆤󠄱󠄾󠄥󠅇󠇕󠆣󠅟󠅀︍󠆼󠄬󠇘󠄯󠆴󠇝󠇌󠅹󠆷󠄴󠄞󠅐󠄊󠆓Document formats had signature capabilities that C2PA could leverage. 󠇟󠇠󠇡󠇢󠅲󠄄󠆇󠄦󠄕󠆚󠄸󠅞󠅹󠇎︅󠇧󠆬󠄈󠄊󠇐󠆊󠇯󠇅︇󠄡󠅻󠇮󠅟󠅊󠅏󠇊󠆖󠅗󠅀󠆚︊󠆈︃󠆛󠄔󠅙󠄪󠆉󠆹### 󠇟󠇠󠇡󠇢󠇬︆󠆏󠇂󠆬󠅨󠄴󠆵󠆉󠅘󠅂󠆘󠇌󠅓󠄣︂󠆔󠄍󠇄󠄯󠆳󠆔󠇉󠆓󠇧󠆂󠄸󠄏󠅎󠇔󠅉󠄶󠇘󠅇️󠅎󠅎󠆎󠆓󠄒What Didn't Exist

Plain text: 󠇟󠇠󠇡󠇢󠄁󠅐󠅵󠆪󠇧󠆬󠄺󠆼󠅿󠆏󠄆󠄘󠅎󠅩󠆮󠅮󠇗󠅌󠅟󠄶󠄰󠆃󠄐󠇠󠅼󠅁︁󠄸󠄈󠇮󠅔󠄮󠅔󠅢󠄐󠆢󠇥󠆅󠇧󠇐No metadata structure. 󠇟󠇠󠇡󠇢󠄁︀󠆫󠆘󠆖󠆺󠄾󠆬󠆉󠇫󠆁󠄕󠆎󠅪󠄓󠄬󠄰󠅩󠆷󠆰󠄦󠅼󠆠󠆁󠅟󠆠󠅭󠅰︅󠇁󠆑󠄐󠇬󠇆󠄤󠄗󠇚󠄹󠅖󠆏No container format. 󠇟󠇠󠇡󠇢󠇍󠅔󠄱󠄀󠅡󠇖󠄸󠆁󠆋󠆖︉󠄁󠆩󠆸󠇟︉󠅓󠅒󠇔󠄣󠆅󠄾︍󠆨󠆿󠅮󠅾󠅈󠆀󠆨󠆆󠅂󠇀󠆪󠅱️󠆧󠅯󠅌󠅇No place to put provenance data. 󠇟󠇠󠇡󠇢󠇤󠅝󠇧󠅛󠄧󠇡󠄰󠇣󠅾󠅶󠄱󠅞󠆼󠇯󠇄󠄞󠅲󠄔󠆒󠆎󠄋󠅲󠆖󠄳󠆫󠇯󠇝󠄕󠇤󠄏󠇯󠆥󠄳󠆩󠆈󠆣󠇟󠅰󠄥󠇎When you copy text from a webpage, what travels with it? 󠇟󠇠󠇡󠇢󠅈󠅾󠅄󠆻󠄟󠇘󠄲︆󠆈󠇪󠇌󠆠︊󠇈󠄘️󠄧󠆰󠇕󠆢󠄢󠆏󠄃󠄬󠄆󠅮󠅧󠄢󠆼󠅷︎󠅮󠅛󠄋󠅕󠅃󠆵󠄻󠆫󠅦Just the characters. 󠇟󠇠󠇡󠇢󠄮󠅿󠇦󠆽󠄘󠄅󠄴︎󠆓󠄝󠅛󠆠󠅭󠄎󠅂󠆋󠅟󠆸󠄬󠄈󠅾󠆤󠅂︃󠆄󠆉󠇚󠆪󠅀󠄙󠆒󠇇󠅱󠄋󠇋󠄆󠇯󠆿󠇘󠄤No author information. 󠇟󠇠󠇡󠇢󠇊󠅕󠇂󠄂︀󠅿󠄼󠇟󠆚󠄢󠆋󠆣󠄧󠆼󠅭󠄎󠆴󠄊󠅺󠅯󠅭󠄠󠄷󠄈󠄒󠅉󠆚󠄔󠅙󠄵󠄡󠆉󠇡󠇓󠄙󠄌󠄍󠆹󠇥󠄥No timestamp. 󠇟󠇠󠇡󠇢󠆇󠄺󠄭󠆝󠇟󠅈󠄵󠄁󠆩󠅝󠆣󠅀󠄓󠄡󠄅󠄸󠆷󠄇󠆬󠆊󠅇󠇕󠇆󠄉󠇐︎󠇎󠅨󠅖󠅵󠄠󠇑󠅸󠆈󠅺󠄁󠆰󠆁󠇄󠄙No proof of origin. 󠇟󠇠󠇡󠇢󠅕󠇆󠇤󠅡󠄠󠆋󠄾󠅽󠆛󠅧️󠄣󠆢󠇆󠄫󠅜󠄆󠅟󠆙󠅯󠇦󠄾󠅴󠄠󠇌󠄜󠄛󠄔󠄅󠆆󠄨󠅫︌󠅤︆󠆡󠆡󠇦󠅃󠆺Nothing. 󠇟󠇠󠇡󠇢󠅤󠆺󠆌󠅏󠅀󠄮󠄲󠄑󠆏󠆅󠄰󠄈󠄸󠇎󠄫󠆳󠅚󠆆󠆳󠆴︀󠅀󠇜󠇩󠆙󠅧󠄤󠄻󠆱󠅙󠄿󠅂󠅉󠅛󠇆󠄝󠄃󠄮󠆮󠆬## 󠇟󠇠󠇡󠇢󠆣󠇪󠅚︅󠆵󠅃󠄵󠆋󠆒󠇛󠆴︇󠅆󠇧󠅟󠇇󠇍󠇫︋󠆄󠅾󠄸󠇭󠅆󠄺󠅘󠄌󠄘︃󠆑󠅦󠆥󠅫󠄁︂󠄕󠅹󠅟󠇌󠄙Why Text Is Different

The challenge with text provenance isn't just technical—it's fundamental to how text works as a medium. 󠇟󠇠󠇡󠇢︂󠆄󠆒󠄀󠆐󠄆󠄾︀󠆉󠇖󠅨󠄛󠆕󠆤󠄨󠄞󠅢󠇚󠆮󠅺󠇇󠇊󠆝󠄳󠅞󠆸󠆨︄󠆹󠅄󠆐󠅶󠅪󠇥󠅘󠇕󠄟󠄭󠄑󠄘### No Container Format

Images are stored in formats (JPEG, PNG, WebP) that have designated spaces for metadata. 󠇟󠇠󠇡󠇢󠇂󠆺󠄷󠆉󠅠󠅀󠄿󠄨󠅴󠆰󠅵󠄠󠅤󠅩󠅮︋󠅷󠆺󠅜󠄹︎󠄛󠆫󠄨︍󠄶󠆉󠆓󠇥󠅐󠅱󠇖󠄈󠄽󠆊󠆀󠆔󠄂󠆊󠄖Text is just... text. 󠇟󠇠󠇡󠇢󠅌󠅘󠄟󠅳󠅐󠄤󠄶󠄃󠅸󠇦󠄴󠆔󠇬︆󠅖︁󠆆󠄙󠇍󠄔󠆍󠆕󠅾󠆍󠄟󠄀󠄭󠅑󠄞󠅾󠄝󠆎󠆪󠅦󠆨󠅾󠄺︌󠅉󠅒A sequence of characters with no inherent structure for additional data. ``` 󠇟󠇠󠇡󠇢󠇖󠆚󠅣󠄿󠄯󠅼󠄶󠅡󠅶󠄦󠇙󠄎󠇬󠆻󠅵󠅳󠆅󠇝󠆙󠅞󠄟󠆚󠄉󠄡󠆶󠆹󠆚󠅟︉󠅃󠆪󠅨󠅑󠄟󠄏󠆉󠆺󠇇󠄨󠄟Image file: [Header][Metadata][Image Data][Footer] Text file: [Characters]


There's nowhere to put provenance information without changing the text itself. 󠇟󠇠󠇡󠇢󠅸󠆖󠅗︊󠄵󠅳󠄱󠆩󠆘︍󠅚󠆅󠆟󠇇󠆡󠄩󠄭󠄟󠅙󠄌󠄪󠆵󠆞󠄜󠅌󠅀󠇙󠆷󠅎󠅞󠆦󠆤󠆐󠆓󠆥󠆡󠄷󠇥󠆰󠆘### Perfect Copyability

When you copy an image, you might lose metadata depending on how you copy it. 󠇟󠇠󠇡󠇢󠅯󠄸󠇌󠄷󠆏󠆀󠄺󠆗󠆫󠆀󠇡󠄆󠅾󠅐󠅇󠆗󠅵󠄜󠅤󠆾󠆥󠆛󠅿󠄪󠆖󠇜󠆈︌󠆷󠄹󠆠󠆩󠇩󠄇󠇭󠄟󠇆󠇂󠇙󠆜But the image itself—the pixels—remain intact. 󠇟󠇠󠇡󠇢󠅸󠄎󠅶󠅽︋󠅚󠄶󠄬󠆙󠅾󠄂󠄧︎󠄛󠅌󠆎󠄪󠆬󠄄󠄎󠆀󠆀󠇞󠅦󠅁󠅿󠄳󠅥󠆀󠄺󠇩󠄥󠄲󠆳󠆺󠆒󠅂󠄲󠆆󠇖Text is different. 󠇟󠇠󠇡󠇢󠆂󠆤󠅞󠆤󠆋󠆅󠄽󠅋󠅲󠇤󠄑󠄂󠇋󠅺󠄌󠆓󠇭󠅋󠅵󠄤󠇅󠄺󠅛󠆔󠆓󠆈󠄎󠇄󠆫󠅇󠆅󠆯󠆓󠅗󠄫󠄟󠄬󠅚󠆭︄When you copy text, you get exactly what you copied. 󠇟󠇠󠇡󠇢󠄜︎󠇙󠄉󠅿󠆆󠄰󠄂󠆗󠆁󠄄󠆩󠆚󠅭󠅠󠇘󠆪︋󠅭︆󠆏󠅽󠄐󠄔󠄩󠆹󠆷󠄵󠆟󠄕󠇞󠅈󠆱︈︁󠄜󠆪󠄊︈󠆃Every character, perfectly reproduced. 󠇟󠇠󠇡󠇢󠆇󠅛󠇙󠇟︎󠆦󠄻󠅽󠅹󠄫󠆀󠅿󠅿󠄭󠄡󠆡󠄸󠇡󠇫󠅙󠄮󠄫󠇅󠆘󠇔󠆤󠄂󠅮󠅒󠆥󠆽󠅟󠅍󠇂󠇓󠄃󠆯󠄻󠇬󠅦This is a feature, not a bug—it's why text is so useful. 󠇟󠇠󠇡󠇢󠆯󠇀󠅃󠄞󠄉󠇬󠄴󠅪󠆨󠅌󠆐󠇢󠅆󠅣󠅈󠇭󠄍󠆵󠇩󠄭󠅘󠄳󠅯󠄱󠅬󠄠️󠅘󠄯󠇄󠆡󠅻󠆗󠄘󠅧󠇯󠆸󠆟󠄄󠅨But it means any provenance solution must survive this perfect copying.

### Format Agnosticism

Text appears everywhere:
- Web pages (HTML)
- Documents (Word, Google Docs)
- Emails
- Chat messages
- Social media posts
- Code files
- Plain text files

A provenance solution that only works in one format isn't really a solution. 󠇟󠇠󠇡󠇢󠅻󠇪󠆨󠆚󠅂󠄑󠄻󠆚󠆝󠇉󠅤󠇟󠆅󠆖󠆍󠆼󠆂󠅆󠇝󠄲󠄃󠆤󠆺󠆗󠄊︉︉󠆉󠄁󠇥󠇒󠅩󠇧󠄨󠅒󠆦󠅊󠅱󠆅󠅤It needs to work everywhere text goes.

### Human Readability

Unlike images, where you can hide data in pixel values, text is meant to be read by humans. 󠇟󠇠󠇡󠇢󠇘󠄻󠇗󠅂󠇃󠆮󠄱󠆷󠅲󠅋󠆛󠄙󠆄󠅈󠅑󠄀󠇅󠄭󠆶󠄝︇󠄡󠄅󠅛󠄲󠅛󠆟󠆳󠄒󠅇󠅏󠇋󠄤󠇑󠅢󠆰󠆠󠇌󠇍󠄢Any embedded data must be truly invisible—not just hard to see, but impossible to see without special tools.

## The Technical Challenge

Given these constraints, how do you embed provenance in text? 󠇟󠇠󠇡󠇢󠆼󠄾󠅹󠆝󠄼󠄶󠄹︇󠆌󠇍󠆈󠄹󠆅󠅸󠅩󠆧󠆖󠄤︁󠆣󠆹󠅷󠅫󠄴󠄘󠅴󠄔󠆀︊󠆆󠄕󠄑󠅭󠄂󠇔󠆳󠄇󠇁󠅻󠅥### Failed Approaches

**Approach 1: Separate Metadata Files**

Store provenance in a companion file (like image.jpg and image.jpg.c2pa). *󠇟󠇠󠇡󠇢󠅳󠆮󠆩󠅠󠆾󠆘󠄳󠆪󠆑󠅍󠇡󠆱󠄸󠄣󠇬󠇐󠅱󠄳󠇪󠄨󠆹󠅼󠇡󠇁󠆠󠅙󠅯󠇒󠆹󠆀󠄚︁󠄈󠅪󠇐󠄇󠅬󠆱󠄺󠇡Problem:* 󠇟󠇠󠇡󠇢󠄦󠄺󠅌︊󠆹︉󠄻󠅰󠅼︁󠆪󠅅󠆅󠇀󠄹󠅠󠅞󠆚󠆚󠅡󠇓󠆸󠆨󠆕󠇚󠇋󠇎󠆁󠅄︆󠆕󠇥󠆣︌󠆝󠅁󠆝󠆿󠆒󠇙The metadata gets separated from the text instantly. 󠇟󠇠󠇡󠇢󠆚󠄈󠆒󠅞󠇖󠇭󠄳󠆾󠅽󠅐󠄄󠅯󠄟󠅑󠇆󠆚󠄭󠅉󠅇󠇏󠅈󠄋󠅖󠄄󠅍︋󠄳󠅊󠆰󠄽󠄠󠆛󠇫󠆳󠄏󠇬󠅖󠆲󠇭󠅊Copy the text, lose the provenance. **󠇟󠇠󠇡󠇢󠅹󠆀󠆕󠇭󠄧󠇃󠄳󠄘󠆭󠆄︄󠄃󠆩󠆌󠆺󠄸󠅳󠇪󠅞󠄆󠅋󠅳󠅚󠅛󠄢󠅀󠆧󠄆󠇗󠅉󠇜󠅑︈󠅤󠄠󠇁󠅨󠄒󠇨󠄰Approach 2: Wrapper Formats**

Create a new format that wraps text with metadata (like a ZIP file containing text and provenance). *󠇟󠇠󠇡󠇢󠆢󠆺󠆣󠅜󠇯󠅷󠄹󠄕󠆐󠅸󠇪󠄾󠇓󠄙󠇦󠅍󠇢󠅥󠇌󠇩󠇜󠅂󠄑󠇟󠆝󠄜󠆚󠆢󠄵󠄵󠆁󠄭󠅃󠆙󠄘󠄸󠅹󠄩︆󠅵Problem:* Requires special tools to read. 󠇟󠇠󠇡󠇢󠇧󠄙󠄡󠆉󠅟󠆅󠄸󠅴󠆂󠄐󠄭󠄪󠄦󠅦󠅏󠇍󠆪󠆔︋󠅻󠅆󠅾󠄁󠅞󠆢󠄬󠇟󠆭󠇃󠄪󠄕󠆀︄󠆟󠆣󠆢󠄸󠇁󠄉󠆔Doesn't work when text is pasted into other applications. **󠇟󠇠󠇡󠇢󠇎󠄶󠄃󠇣󠆟󠅢󠄰󠄩󠆛󠆨󠆺󠄜󠄗󠇃󠅡󠇫󠇘󠅕󠇇󠅗󠅻󠅃󠄕󠇣󠇟󠅁󠅳󠇢󠅢󠇡󠄥︌󠄿󠇖󠅂󠄃󠆍󠄚󠅈󠄆Approach 3: Visible Markers**

Add visible attribution text ("© 2025 Publisher Name"). *󠇟󠇠󠇡󠇢󠇋󠅥󠆣󠄷󠇩󠆅󠄴󠇢󠆨󠇍󠅜󠅳󠇏󠆦󠇒󠇯󠅑󠅳︉󠇕󠄲󠇨︁󠄮󠅟︄󠄫󠆍󠄖󠆯󠆮󠆞󠄡󠅂󠆨󠇯󠆒󠇏󠅱󠅙Problem:* Easily removed. 󠇟󠇠󠇡󠇢󠄹󠅩󠆞󠅽󠄨󠇊󠄲󠆐󠆛󠇑󠄆󠅱󠆦󠄵󠄹︃󠆿󠅛󠄘󠅨󠆓󠆊󠅱󠆺󠆘󠆏󠄵󠇯󠅔󠄔󠄐󠆏󠄅󠅚︅󠇤󠅌󠅑󠆜󠆌Doesn't provide cryptographic proof. 󠇟󠇠󠇡󠇢󠆋︃󠆬󠄆󠆽󠇌󠄾󠆆󠆠󠆈󠅗󠄠󠅄󠇖︊󠅚󠇐󠅬󠄳󠅋󠆿󠅞󠆋󠅻󠄙󠄛󠆛󠆝󠄷󠄺󠄢󠅐󠅯󠄇󠅘󠇦󠇞󠄜󠅹󠅵Changes the content. **󠇟󠇠󠇡󠇢󠆅󠇒󠄖󠅝󠆲󠅉󠄻󠅒󠆞󠅈󠆨󠅤󠆳󠅂󠆛󠆖󠄐󠇉󠆟󠄓󠆼󠆚󠄢󠇓󠆨󠆏󠄱󠇆󠅹󠅎󠆱󠅾󠅏󠄞︎󠇗󠅋󠄷󠅕󠄓Approach 4: Steganography** Hide data in the text using techniques like varying word spacing or using homoglyphs (similar-looking characters). *󠇟󠇠󠇡󠇢󠄁󠄴󠆖󠄳󠆖󠆞󠄱󠄬󠅵󠅁󠅴󠆗︄󠄔󠆒󠆭󠅳󠄪󠇣󠆀󠇙󠄋󠇠󠆗󠄚󠅒󠆤󠆲󠅞󠄶󠇘󠆟󠇥󠅮󠇎󠇆󠄠󠆜󠄉󠄕Problem:* Fragile—doesn't survive reformatting. 󠇟󠇠󠇡󠇢󠄻󠆚󠆖󠇕󠅼󠄷󠄽󠆹󠆠󠄄󠄉󠇈󠇉󠅌󠅬󠆝󠄘󠅈󠄡󠅚󠇐󠅫󠆖︋︉󠆯︇󠅗󠄛󠇝󠅃󠅴󠇋󠆃󠅲󠄑󠅟󠆅󠄊︍Detectable and removable. 󠇟󠇠󠇡󠇢󠅣︎󠅤󠅫󠄧󠇬󠄱󠅬󠆇󠇉󠄺󠆤󠆦󠇙︋󠄇󠅧󠄞󠇦󠅔︆󠅶󠅡󠄕󠄧󠇚󠅨󠅳󠆊󠅟󠄔󠄶󠄣󠆎󠅉󠆷󠅛󠅤󠄄󠇑May cause rendering issues. 󠇟󠇠󠇡󠇢󠇫󠄬󠅊󠇓󠅉󠄧󠄽󠄽󠆔󠆰󠅔󠆍︌󠆒󠆮󠆨󠅒󠅼󠄻󠅳󠅹󠇪󠆛󠄯󠄐󠄞󠆥󠇖󠆉󠄋󠇪󠆾󠄫󠄛󠅼󠆖󠄻󠇝󠄍󠅾### The Unicode Solution

The breakthrough came from an unexpected place: Unicode variation selectors. 󠇟󠇠󠇡󠇢󠄿󠇍󠆪󠇅󠄃󠅃󠄱󠄜󠆩󠆯󠄖󠅴󠆐󠇙󠆣󠅒󠄘󠅄󠄻󠅓󠅣󠅷󠅼󠅙󠆲︂󠇋︀󠅬󠄎󠆾󠅛󠇒󠇄︄󠅩󠄲󠄎󠇩󠆋Unicode includes characters called "variation selectors" (U+FE00 through U+FE0F and U+E0100 through U+E01EF) that are designed to modify how the preceding character is displayed. 󠇟󠇠󠇡󠇢󠆱󠅟󠅱󠆶󠄙󠇓󠄸󠅕󠆌󠅯󠇐󠆷󠆅󠆁󠅱󠆧󠇙︇󠄊󠆷󠅄󠇃󠆠󠄢󠄌︈󠅕󠇢󠆙󠅾󠇔󠆼󠄮󠄆󠄬󠅘󠄏󠄒󠅆󠅅In practice, most text systems ignore them—they're invisible and have no effect on rendering. 󠇟󠇠󠇡󠇢󠆨󠆌󠅬󠇡󠄓󠆈󠄻󠄔󠆉󠆍󠇇󠅤󠄰󠄹󠇘󠅌󠇅󠇘󠆡︃󠇫󠄈󠇤󠅲󠄊︄󠆩󠇤󠇊󠅳󠆕󠆫󠇮󠄦󠅉󠄥󠅌󠇊︇󠄼This creates an opportunity: you can embed data using these invisible characters without changing how the text appears or behaves. ```
󠇟󠇠󠇡󠇢󠆌󠆌󠄼󠇅󠆀󠇏󠄸󠆈󠅸󠄘󠄹󠆫󠇑󠄤󠅯󠆂󠅢󠄓󠇃︃󠅈︊󠄘󠇏󠆟󠄴󠄰󠅼️󠅎󠅨󠇋󠇍󠅵󠅂󠅊︈󠄎󠄫󠇐Visible text: "Hello, world!"
With embedded data: "H[VS]e[VS]l[VS]l[VS]o[VS], world!"
(where [VS] represents invisible variation selectors encoding data)

The text looks identical. 󠇟󠇠󠇡󠇢󠇠󠄝󠅅󠇗󠇟󠆣󠄻󠇇󠆜󠄷󠆲󠅶󠅿󠇙󠄘󠄸󠆝󠆇󠆪︉󠄐󠅽󠄼󠅰󠇊󠇢󠄑󠅇󠄫󠄨︅󠅄󠅇󠇌󠅶󠅎󠇊︆󠄳󠄞It copies identically. 󠇟󠇠󠇡󠇢󠆞󠆥󠅗󠇋󠆐󠄆󠄾󠆘󠆂󠆎󠇊󠄍󠄘󠇐󠄌󠇑󠄯󠄇󠆼󠅮󠅪󠇤󠅗󠆗󠆏󠄿󠄷󠇑󠅸󠆗󠅦󠅆󠇏󠇙󠅂︌󠇅󠄇󠄖󠄘It renders identically. 󠇟󠇠󠇡󠇢󠅣󠇈󠆏󠅢︊󠄃󠄷󠅔󠆪󠇘󠄟󠆬󠄿󠅢󠅖󠆻󠄵󠆦󠅘󠄌󠇡󠇇󠇘󠇗󠇆󠇄󠆄󠅹󠇒󠄈󠅣󠆫󠇎󠄡󠄯󠇧󠄒󠅩󠆮󠅜But it carries hidden data that can be extracted and verified.

󠇟󠇠󠇡󠇢󠅬󠄴󠆗󠇚󠇌󠇆󠄺󠇪󠆑󠅐󠆘󠇃󠇔︅󠆴󠆴󠄆󠅍󠅤󠆆󠅹󠆀󠅍󠄐󠆽󠅈󠆤󠆳󠅦󠅥󠄇󠇒󠅆󠅑󠅀󠄤󠇠󠅀󠅨󠆶The C2PA Text Specification

As Co-Chair of the C2PA Text Provenance Task Force, I worked with colleagues from Google, BBC, OpenAI, Adobe, and Microsoft to develop the official specification for text provenance.

Section A.7: Embedding Manifests into Unstructured Text

The specification, published in C2PA version 2.3, defines:

The C2PATextManifestWrapper Structure:

Magic number for identification: C2PATXT\0
Encoding using variation selectors
Prefix character (Zero-Width No-Break Space) for compatibility
Validation rules for extraction

Key Design Decisions:

󠇟󠇠󠇡󠇢︁󠄥󠆆󠅍󠄘󠄖󠄵︊󠆙󠇩︉󠇞󠆒󠆺󠆖󠇜󠇑󠆞󠅦󠄽󠄙󠅾󠆙󠅀󠅧󠅕󠄂󠆜󠇞󠆧󠆛󠄧󠆦󠅔󠇜󠄬󠇪󠆧󠄋󠄬Invisibility: 󠇟󠇠󠇡󠇢󠆍󠄅󠇊󠄟󠇗󠆉󠄿󠆞󠆁󠅙󠇆󠇝󠅼󠅻󠇣󠅀󠆮󠄦󠆵󠆃󠅥󠄐󠇚󠇄󠅘󠅷󠆬󠆔󠅨󠇄󠇍󠆦󠅖󠄜󠇮󠆻󠄮󠅫󠆋󠆑Embedded data must not affect text rendering
󠇟󠇠󠇡󠇢󠇤󠄱󠅓︍󠆈󠆴󠄳󠄝󠅳󠄙󠆳󠇞󠄏󠄏󠇖󠄦󠆖󠄟󠅄󠆒󠇝󠅫󠆺󠄙󠄤󠅦󠄠󠇟︍󠆃󠆯󠆞󠇣︃󠅗󠅌󠄘󠅩󠆶󠇪Survivability: 󠇟󠇠󠇡󠇢󠇗󠄅󠇆󠆓󠄮󠄀󠄱󠇜󠅰︋󠇝󠅹󠅓󠅅󠇗󠆊󠇉󠇄󠅢󠄃︋󠆍󠇥󠅦󠅟󠆰󠅖󠄗󠆰󠄃︁󠇪󠇞󠇉󠅻󠆾󠅸󠇅󠇖󠇆Data must survive copy-paste operations
Verifiability: Cryptographic signatures enable proof of origin
Interoperability: 󠇟󠇠󠇡󠇢󠅡󠆐󠇬󠅂󠇧󠆟󠄳󠇎󠆙󠅿︈󠄺󠅤󠄢󠅶󠆘󠇏󠇇󠅣󠆭󠄡󠇥󠇆󠄗󠆄󠆾󠇕󠇫󠅉󠅲󠆖󠆂󠆍󠄓󠇫󠅜󠆌󠆕󠆍󠅓Works with existing C2PA infrastructure
Efficiency: 󠇟󠇠󠇡󠇢󠇎󠅛󠆡󠄓󠇆󠆄󠄽󠇤󠅼󠄋󠇙󠇁󠅢󠄟󠅬󠆂󠅽󠇁󠆨󠅺󠄪󠅵󠆛󠇇󠇉󠅘󠄝󠅣󠇖󠅶󠄟︆󠄖󠆧󠇣︁󠇖󠄐󠆃󠇭Minimal overhead for embedded data

What Gets Embedded

A text provenance manifest includes:

Claim Generator: 󠇟󠇠󠇡󠇢󠆏󠅇󠆂󠆲︁󠄞󠄿󠅉󠅺󠆔󠇨󠇮󠅼󠆩󠇩󠅇󠆶󠆻︈󠆘󠄸︁󠆟󠄛󠅲󠄝󠄬󠆓󠅅󠄣󠆐󠆊󠇧󠇇󠄭󠆆󠅫󠇬󠆭󠆟What system created the content
Assertions: Claims about the content (authorship, creation date, AI involvement)
Signature: 󠇟󠇠󠇡󠇢󠇫󠇞󠇦󠅃󠅆󠅈󠄶󠅄󠆯󠅿󠅣󠅲︂󠇁󠄋󠄛󠇢󠆖󠅳󠆁󠆠󠆕󠄙󠆳󠄬󠆍󠄾︃󠆸󠆾󠆥︋󠅰󠆋󠆼󠅣󠆣󠇐󠅜󠇨Cryptographic proof binding assertions to content
Hash: Integrity check for the text content

󠇟󠇠󠇡󠇢󠄼󠅴󠇭󠆂󠆁󠇏󠄻󠆘󠆬󠅦󠅦󠄺󠅣󠇢󠇫󠄑󠄮󠆙︍󠅝󠆧󠄡󠄱󠆿󠄚󠇎󠆼󠇀󠄤󠆶󠆼󠄲󠅸󠇞︍󠆖󠇒󠇓︍︍Why This Matters Now The timing of text provenance isn't coincidental. 󠇟󠇠󠇡󠇢︎󠅬󠇠󠆓󠄺󠇗󠄸󠆾󠆓󠅎󠇓󠄘󠄰󠇦󠅤󠆫󠅴󠅊󠄓︈󠄚󠆘󠄠󠅩󠅧󠄤󠅵󠆸󠆯󠆳󠅹󠆼󠅅󠅙󠄚󠅲󠆻󠆉󠆃󠄉Several factors converged to make it essential:

AI-Generated Text Explosion

The vast majority of AI-generated content is text:

ChatGPT responses
AI-written articles
Automated emails
Code generation
Content at scale

Without text provenance, there's no way to verify the origin of this content.

Regulatory Requirements

The EU AI Act and California SB 942 require machine-readable marking of AI-generated content. 󠇟󠇠󠇡󠇢󠅒󠆅󠅗󠆄󠅑󠅬󠄴󠇃󠆕󠅟󠆉󠄾󠇁󠇮󠅟󠇒󠅘󠄘󠅍󠇘󠆞󠆌︇󠇩󠄑󠅣󠅝󠄘󠄨󠄢󠇃󠅍󠇯󠆶󠄜󠄭󠅹󠇡󠆃󠅈For text, this was impossible without a provenance solution. 󠇟󠇠󠇡󠇢󠅂󠆢󠅼󠆳󠆻󠆉󠄲󠆔󠅱󠇢󠄗󠆋󠄕󠆮󠇤󠆩󠅆󠆒󠅬󠄯󠅡󠅖󠆌󠅶󠇕󠄊󠄝󠇊󠇄󠆈󠅻󠄺󠆫󠅋󠄶󠅜󠄯󠅤󠅰󠄬### Publisher Needs

Publishers facing AI training challenges need to prove ownership of their content. 󠇟󠇠󠇡󠇢󠄳︀󠄷󠅈󠇇󠄷󠄽󠅃󠆙󠆼󠅶󠅬󠆥󠄓󠄇︍󠄕󠅉󠅓︇󠇈󠄐󠅪󠇑󠇃︊󠅻︄󠆡󠅫󠆈󠅂󠆭󠆊󠆵󠅉󠆍󠇕󠇐󠆿Text provenance enables:

Proof of publication date
Cryptographic ownership verification
Tracking through distribution chains
Licensing enforcement

󠇟󠇠󠇡󠇢󠇙󠆬󠄔󠅇󠇋󠄹󠄾󠆔󠆯󠅹󠆠󠇮󠅌︎󠆣󠆂󠇦󠅗󠆹󠆴󠅾︈󠆅󠇎󠇊󠆒󠅖󠅁󠇤󠇛󠆑󠅭󠆻󠄘󠆗󠆱󠆚󠇒󠇀󠆅The "We Didn't Know" Problem

AI companies have claimed they couldn't identify specific content among billions of documents. 󠇟󠇠󠇡󠇢󠄠󠄟󠆩︋︆󠇇󠄶󠆈󠆘󠅕󠅜󠆡󠄮󠇓󠆊󠅐󠆉󠄥󠄖󠅩󠇓󠆑󠅩󠅭󠄸󠄽󠄾󠄈󠄋󠄝󠇩󠄬󠄌󠆼󠄮󠇨󠅂󠆋󠆪󠇕Text provenance eliminates this defense by making content self-identifying. 󠇟󠇠󠇡󠇢󠇯󠆻️󠆅󠇒󠄅󠄻󠅀󠆄󠆠󠄘󠄗󠄨󠇒󠇐󠆯󠅄󠄢󠅠󠆧󠇬󠄦󠇥󠅉󠅱󠅓󠅪󠅃󠄔󠆴󠄨󠄃󠅻󠆛󠅕󠇛󠄄󠄰󠄤󠇕## 󠇟󠇠󠇡󠇢󠇫󠅠󠅻󠄧󠅐︁󠄷󠄁󠆔󠅜󠅴󠅂󠅖󠄰󠄍󠅺󠄔󠇬󠆼󠅮󠆎󠄼󠅩󠆎󠄾󠄍󠅣️󠄆󠄃󠅽󠄗󠄕󠇇󠄬󠄉󠅳󠅁︍󠅨What Text Provenance Enables

With the technical foundation in place, new capabilities become possible:

For Publishers

Proof of Origin: 󠇟󠇠󠇡󠇢󠅚󠄾󠆆󠇖󠄂󠆔󠄽󠅠󠆟󠆚󠇬󠆕󠆞󠆱󠅙󠅴󠄲󠆹󠅊󠅳󠆘︈󠇙󠄻󠄣󠇍󠅈󠆤󠆷󠄽󠄃󠄿󠅆󠄓󠅇󠇩󠅫󠅐󠇞󠆛Every article carries cryptographic proof of who published it and when. 󠇟󠇠󠇡󠇢󠅵󠄭󠅨󠅕󠇎󠇎󠄽︂󠅺󠆚󠇎󠄉󠅮󠄟󠅾󠇓󠄢󠄿󠇋󠆦󠅰󠅺󠄗󠆀󠆔󠆱󠆐󠆩󠅇󠄹󠅔󠇅󠇊󠅈󠆜󠄖󠄱󠆚󠅢󠇎Downstream Tracking: Content remains identifiable as it flows through syndication, aggregation, and scraping. 󠇟󠇠󠇡󠇢󠄭󠄎󠆼󠅵︍󠄟󠄶󠇚󠆌󠅋󠇍󠆨󠅸󠄘󠅸󠇙󠆵󠄛󠆒︅󠅀󠄒︄󠇈󠇘󠅨󠅍󠇕󠅐󠆈󠆣󠇗︄󠅹󠇖󠅈󠅔󠅤󠄾󠅹Licensing Enforcement: 󠇟󠇠󠇡󠇢︉󠆌󠆋󠆩󠅖︃󠄵󠄜󠆠︇󠄙󠅾󠄩󠇌󠅪󠆾󠆦󠄏󠆋󠄅󠇨󠇓󠆞󠆔󠅚󠄾󠅢󠇙󠅚󠅟󠆮󠅞󠅝︀󠄽󠅷︈󠅆󠇗󠆨Unauthorized use can be detected and proven. 󠇟󠇠󠇡󠇢󠄮󠆟󠆲󠅔︎󠆤󠄿󠅷󠆟󠇐󠇐󠄗󠆋󠇋󠄄󠆖󠄰󠆉︍󠆝󠇠󠇐󠆊󠅃󠇣󠄚󠆅󠆸󠄖󠆖󠆿󠇀󠄫󠅺󠆣󠇬󠆶󠆖󠇕󠇯Quote Integrity: Verify whether AI attributions match your actual content. 󠇟󠇠󠇡󠇢󠄰󠅓󠇊󠆆󠆁󠄬󠄾󠅔󠆖󠄨󠄪󠇐󠄖󠄉︍󠇈󠇔󠇘󠆵󠄎󠆂󠅑󠇊󠄡󠇟󠅞󠆲󠄵︂󠄕󠇗󠆑󠆵󠄍󠄞󠆈󠄺󠇜️󠄙### For AI Companies

Content Verification: Identify the source of training data. 󠇟󠇠󠇡󠇢󠇀󠆘󠇡󠆳󠅰󠄙󠄰󠆯󠅽󠆃󠅦󠇘󠄱󠆻󠆑󠆝󠆚︁󠅕󠇒󠇮󠆖󠄥󠇧󠅏󠄛󠆩󠅼󠇢󠄋󠆛󠄂󠆃󠆐󠅜󠄭󠄙󠅟󠄼󠇅Compliance: 󠇟󠇠󠇡󠇢󠅳󠇖󠆜󠆞󠇖󠆛󠄺󠅝󠆟󠆂󠄕󠅒󠆧︁󠄨󠇣󠄕󠆣󠅜󠅳󠄿󠅡󠄮󠆵󠄒󠅡󠇅󠇬󠅀󠇙󠅔󠇮󠇡󠆍︊󠅑󠅩󠇋󠆎󠇄Meet regulatory requirements for content transparency. 󠇟󠇠󠇡󠇢󠇈󠇗󠆣󠄎󠄿󠄴󠄷󠆚󠆣󠆊︁󠄥󠅺󠅕󠇩󠇀󠄋󠅣󠅚󠆛󠅫󠇙󠅆󠄟󠆆󠅄󠄕󠄗󠆅󠇎󠇪󠅄︂󠄪󠆇󠅎󠅖󠆃󠄿󠅳Quality Assurance: Verify human-created vs. AI-generated content. 󠇟󠇠󠇡󠇢󠆈󠆶󠄻󠆨󠆇󠇧󠄻󠄡󠆨󠅀󠆈󠇖󠆡󠅾󠆽󠄘󠅛󠄃󠇇󠇃󠅣󠆲󠆤󠅡󠆓󠆐︃󠄇󠄰󠅴󠅽󠄽󠄧󠇠󠅇󠇌󠄯󠆾󠆷󠆊Attribution: 󠇟󠇠󠇡󠇢󠇨󠄉󠆦󠇑󠆾󠄬󠄹󠅊󠆜󠆊󠅇󠄯︊󠄙󠄌󠅺󠅮󠄫󠅲󠆢󠅁󠆀󠆢󠆫󠄽󠄟󠇧︉󠇂󠅧󠆤󠆀󠆴󠆶󠅒󠇌󠅫󠅊󠅈󠇉Properly credit sources in AI outputs. 󠇟󠇠󠇡󠇢󠆄︅︁󠅀󠅣󠅤󠄹󠄮󠆟󠅇󠅍󠄖󠄢󠄾󠆏󠆔󠅧󠄿︎󠆴󠇥󠆣󠇤󠆇︍󠇒󠆙󠆕󠅻󠅙󠆩󠆕󠅊︄󠇭󠅨󠄅󠅆󠅷󠆑### For Users

Trust Verification: Check the origin of content you're reading. 󠇟󠇠󠇡󠇢󠅈󠇤󠅷󠅇󠆋󠆉󠄰󠆚󠆂󠇇󠇔󠆓󠇇󠄼󠇂󠅑󠄩󠇅󠄄󠅽󠆨󠇡󠅱󠅀󠅺󠄊󠆅󠆿󠆈󠄹󠆉󠇩󠄴󠇇󠆝󠇕󠇧󠇈󠆟󠅣AI Detection: Know whether content was AI-generated. 󠇟󠇠󠇡󠇢󠆮️󠆓󠇑󠆉󠄡󠄴󠆿󠆊󠅭󠆤󠅢󠄷󠆞󠆣󠄏󠇗󠇓︈󠅵󠅢󠅴󠆚󠄚︈︋󠅦󠄽󠄪󠆝󠄃󠄔󠆎󠅆󠅿󠆁󠆟󠅇󠄌󠇑Source Validation: Verify attributed quotes and claims. 󠇟󠇠󠇡󠇢󠆧󠄸󠆵󠆝󠇋󠆻󠄾󠇛󠅺󠇭󠆻󠆚󠇑󠇓󠄋󠄆󠆸󠄯󠅼󠇤󠇕󠆨󠄐󠄳󠆌︍󠇜󠄗󠅃󠄐󠆖󠆹󠇉󠅢󠇯󠇢󠇚󠅢︀󠇢## The Road Ahead

Text provenance is now technically possible and standardized. 󠇟󠇠󠇡󠇢󠄭󠅁󠇥󠇏󠄽󠆏󠄵󠄩󠆁󠅾︇󠆧︄󠆁󠄔󠆪︆󠇃󠄮󠆐󠅒󠇊󠆅󠅡󠆅󠅴︁󠇓󠆼󠇉󠇤󠆙󠆃󠅁󠆹󠆘󠆒󠄀󠆥󠄴The next phase is adoption:

Infrastructure Buildout

Publishers implementing signing at publication
AI companies integrating verification
Platforms supporting provenance display
Tools for end-user verification

Ecosystem Development

Licensing frameworks built on provenance
Enforcement mechanisms using cryptographic proof
Market standards for verified content
Regulatory compliance pathways

Continued Innovation

Sentence-level tracking for granular attribution
Quote integrity verification at scale
Performance intelligence from attribution data
New applications we haven't imagined yet

Closing the Loop

For years, content authenticity had a text-shaped hole. 󠇟󠇠󠇡󠇢󠅽󠄰󠅟󠄸󠆄󠅾󠄻󠇮󠆛󠆔󠅦󠅥󠆟󠅜󠅼󠅅󠆯󠄃󠅜󠄲󠅦󠆋󠆺󠆳󠅽󠄈︊󠇔󠆱󠆭󠇣󠆚󠆑󠅲󠄴󠆃󠅄󠄕󠅅󠆫Images, video, audio, and documents all had provenance solutions. 󠇟󠇠󠇡󠇢󠆷󠄎󠇧︉󠅾︉󠄻󠅎󠆡󠅣󠄸󠇧󠆜󠇖󠄮󠇏󠆇󠅦󠇧󠄢󠇛󠇒󠇫󠅨󠄍󠄽󠆞󠇛󠆥󠆙󠇠󠅔󠇂󠄘󠇗󠆛󠄃󠄦󠇞󠄝Text—the most fundamental and ubiquitous form of digital content—had nothing. 󠇟󠇠󠇡󠇢󠆱󠄫󠆱󠅽󠇀󠄟󠄽󠄌󠆜󠅇󠄱󠇯󠆡󠇄󠄈󠅥󠆐󠄵󠅬󠅔󠅊󠄶󠄐󠇍󠆌󠄌󠅻󠅣︃󠅟󠇁󠅼󠆭󠄉󠆅󠇊󠇪󠇡󠄩󠇉That gap is now closed. 󠇟󠇠󠇡󠇢󠄄󠆇󠄑︂󠄣󠆞󠄷󠇋󠆋󠇘︁󠅰󠆻󠆙󠇚󠆸󠅺󠄤󠅩󠄁󠇄󠆇󠇒󠆉󠄶󠅈︅󠄕󠄕󠄨󠆝󠄁󠅖󠄩󠄢󠄰󠄪󠆲󠇛󠅌The C2PA text specification provides the technical foundation. 󠇟󠇠󠇡󠇢󠇟󠇔󠅣󠇠󠄭︃󠄱󠆁󠆏󠄙󠇪󠅜󠆝󠇇󠆁󠄏󠇩󠇚󠄶︈󠄥󠅒󠅷󠄉󠄍󠆔󠇇󠅩󠄆︆︊︈󠆧󠄬󠄓󠇀󠄢󠆡󠇃︋Implementations are production-ready. 󠇟󠇠󠇡󠇢󠄣️󠅴󠆤︎󠆼󠄺󠅒󠆢󠄸󠇛󠅬󠇂󠄾󠅁󠆁󠅊󠅸︊󠆥󠆺󠅏󠇀󠆹󠅨󠄱󠆐󠆈︅󠆊󠄻󠇫󠇛󠆻󠆤󠅜󠆜󠆙󠆣󠅆The ecosystem is forming. 󠇟󠇠󠇡󠇢󠄻︎󠅾󠇧󠅙󠆳󠄿󠅜󠆒󠇉󠅝󠄈󠇥󠆞󠆪󠅰󠅚󠇫󠆀󠄼︆󠅋󠆣󠆭󠇂󠆃︈󠄤󠆮󠆒󠅺󠄺︅󠅮󠇙︆󠆅󠅆︌󠄮Text was the missing piece. 󠇟󠇠󠇡󠇢󠆌󠆬󠆍󠇚󠆒󠆅󠄰󠇂󠆮󠆽󠅖󠄠󠄛󠄳󠇑󠆚󠅗󠄼︌󠆾󠄒󠄂󠆢󠅱󠆍󠄫󠇡󠄶︃󠄦󠆞󠆎󠄶󠄩󠄩󠅹󠄯󠇯󠇅󠇀It's not missing anymore. 󠇟󠇠󠇡󠇢︅󠆋󠇖󠄉󠄚️󠄻󠄤󠆋󠆞󠅱︉󠆫︍󠆬󠅕󠅼󠄒󠄕󠄂󠄺󠇦󠅅󠆑󠄴󠇔󠄔󠅢󠅾󠇡󠅚󠄿󠇑󠄗󠅥󠆝󠆈󠅇󠆍󠇈---

The standard is published. 󠇟󠇠󠇡󠇢󠆱󠅑󠅎󠇐󠅱󠅏󠄺󠆴󠆐󠅟󠇤󠆁󠄾󠅫󠆪󠆇󠇚󠅰󠅛󠄆󠄍󠄊󠇆󠇦󠆡󠇦󠇗󠅅󠄊󠅣󠅤󠅆󠆌󠄸󠄣󠄂󠄾󠇯󠇤󠆲The infrastructure is ready. 󠇟󠇠󠇡󠇢󠅴󠄌󠄊󠆻󠆙󠅅󠄾󠄚󠆆︆󠇅󠇞󠆣󠄙󠅧︃󠇢󠄺󠇝󠅄󠅫󠇔󠇣󠆿󠇇󠄨󠆌󠅇︄󠆜󠇓󠄲󠆧󠄢󠇣󠄙󠆩󠅩󠆎󠄊The question now is who will lead in implementing it—and who will be left explaining why they didn't. 󠇟󠇠󠇡󠇢󠅎󠅷󠆌󠇡󠄙󠅷󠄽󠅯󠆄󠅯󠄉󠆈󠄌󠆮󠆿󠇮󠄣󠄚󠄭󠆍󠆳︍󠅻󠇚󠄺󠇥︇󠆘󠇖󠄕󠇌󠇋󠆼︌󠅕󠆮󠄔︆󠅝󠄷Learn more about text provenance: 󠇟󠇠󠇡󠇢󠅳󠆈󠄂󠆧󠆥󠄕󠄱󠅟󠆊󠇋󠄐󠇫󠆎󠅱󠆈󠇙󠇑󠆜󠆶󠄨󠄳󠅗󠄀󠅲︃󠅇󠅒󠇋󠆇󠅣󠇐󠅤󠄆󠄮󠆯󠆝󠅐󠇃󠅗󠅲encypherai.com

Read the C2PA specification: 󠇟󠇠󠇡󠇢󠅑󠄏󠅻󠆇󠆣󠆌󠄲󠄔󠆋󠄒󠇞󠆺󠆂󠇗󠅔󠇫󠇂󠆏󠅳󠅑󠄰󠅢󠆶󠄼󠅄󠆵󠄖󠄺󠅉󠄙󠆐󠇓󠅰󠄊󠆞󠇍󠆗󠇛󠇝󠆽Section A.7 - Embedding Manifests into Unstructured Text

#TextProvenance #C2PA #ContentAuthenticity #Innovation #Standards󠇟󠇠󠇡󠇢󠅈󠆯󠅗󠆾󠇐󠇀󠄹󠅍󠆓︈󠆊󠆝󠅏󠄐󠇡󠄇󠄡󠆕󠆽󠄯󠄼󠄎︄󠅿󠅷󠆙󠅍󠆅󠄳󠆋󠆣󠆃󠄎󠆳󠄬󠅃󠆭󠅖︉󠆞