Spymarks, Not Watermarks

A "spymark" is an emerging development of the traditional "watermark," embedded in digital and physical media to trace and track content without the creator's knowledge or consent. Google's SynthID is an example, embedding secret signals into images, audio, text, and video that can encode database identifiers linked to a user's identity.

These identifiers may include full names, IP addresses, date of birth, physical addresses, and political party affiliations. SynthID-O can encode a 136-bit payload in a 512x512-pixel image, with 72 bits left for error correction.

Other tech companies, including OpenAI, are also developing spymarking systems. These systems claim to help identify AI-generated content but go beyond watermarking to build robust tracking mechanisms. As spymarking algorithms proliferate, they may be integrated into social media, content production tools, and smartphones, leading to pervasive surveillance.

Spymarks differ from traditional watermarks, which are visible and removable. Spymarks are hidden, invisible, and often resistant to removal or editing. They encode user-tracking identifiers that are opaque and useless to the user. Moreover, these signals can survive metadata removal and some edits, ensuring that marked copies remain traceable.

The implications of spymarks are concerning. They can be used to track individuals, especially whistleblowers or those with dissenting views. The future may involve every device being attested and every social media post carrying an account-linked spymark. This could make it dangerously easy to track the flow of ideas and identify those who promote "dangerous" viewpoints.

To raise awareness about this issue, the author suggests using the term "spymark" instead of "watermark." This change of terminology puts the privacy concern front and center, making it clear that spymarks are a form of metadata that users have limited knowledge and control over. The privacy risk inherent in spymarks should be recognized and addressed to prevent a dystopian future where personal data is embedded in every piece of media we create and consume.