Data & Privacy
The Cryptographic Birth Certificate for Every Image
The deepfake problem is not, at its root, a detection problem — it is a provenance problem, and the most important attempt to solve it is already embedded in your next camera.
The Idea
For decades, the implicit contract of a photograph was simple: seeing was believing. That contract is broken. AI image and video generators now produce content that is, under ordinary viewing conditions, indistinguishable from authentic recordings — and the gap is widening faster than detection tools can close it. The tempting response is to build ever-smarter detectors. But detection is an arms race with no finish line: every improvement in synthesis outpaces the forensics trained against yesterday's models. The more durable answer is provenance — establishing an unbroken chain of trust from the moment content is created, rather than trying to authenticate it after it has circulated. This is the logic behind C2PA, the Coalition for Content Provenance and Authenticity, a cross-industry technical standards body established in February 2021. The mechanism is elegant: at the point of capture or generation, software or hardware embeds a cryptographically signed manifest into the file — a kind of birth certificate recording who made it, when, with what tool, and what edits followed. The signature is tamper-evident: alter the image and the manifest breaks. Alter the manifest and the broken signature shows up under verification. Crucially, provenance does not ask whether content is true. It asks something narrower and more answerable: does this file's history match what it claims? That reframing is important. A signed credential from a known camera is strong evidence of authenticity. The absence of one is not evidence of fakery — most legitimate content still carries no manifest at all. What provenance gives you is a floor of verified trust, not a ceiling of perfect detection.
In the World
In late 2023, Leica released the M11-P — an expensive rangefinder aimed squarely at photojournalists — notable for one genuinely novel feature: every photograph it takes is automatically signed with a cryptographic manifest at the moment of capture. The camera embeds metadata including the device, lens, date, time, and location in a C2PA manifest, then signs it with a private key stored in a secure chipset. The result is a verifiable provenance record that travels with the file wherever it goes. Leica's market share is tiny, but its significance was conceptual: it demonstrated that cryptographic provenance could be embedded at capture without slowing the camera down or noticeably increasing file sizes. What followed illustrated both progress and fragility. Nikon added C2PA support to the Z6 III via firmware in August 2025, then suspended the service weeks later after a researcher demonstrated that the camera could be tricked into signing images it had not actually captured; all certificates were revoked, and the service remained offline into 2026. On the generative side, the picture is clearer: Microsoft has confirmed that all images created with Copilot carry content credentials based on the C2PA standard, Adobe applies them across its Firefly tools, and OpenAI has extended C2PA metadata to images generated through ChatGPT and the API. The broader ecosystem is organised across two distinct bodies: C2PA is the technical standards organisation, while the Content Authenticity Initiative — the Adobe-led community that promotes adoption — reported more than 6,000 members by early 2026, spanning technology companies, camera makers, newsrooms, and government bodies.
Why It Matters
The practical upshot is not that you need to understand cryptographic manifests. It is that the information landscape is quietly bifurcating: content that can prove where it came from, and content that cannot. For now, most content — including most authentic photographs — falls into the second category. But the direction of travel is clear, and it matters for anyone who produces, publishes, or consumes media. For journalists and publishers, C2PA-enabled workflows are becoming a credibility signal — a way to offer readers something beyond 'trust us'. For anyone sharing visual evidence of anything contested — a protest, an accident, a political event — a cryptographically signed file carries a kind of weight that an unsigned one simply cannot. The Nikon episode is worth sitting with, too. It showed that provenance infrastructure is only as strong as its weakest implementation: a signing vulnerability didn't just affect future images, it retroactively voided the trust signal on every photo those cameras had already signed. The larger shift is real and growing — but it is still, in 2026, unevenly distributed across devices, platforms, and habits, and the chain has to be complete to mean anything.
A Question to Ponder
If provenance credentials become the standard way of vouching for authentic media, what happens to the credibility of everything — the vast majority of content — that simply has no credential at all?
One reader's Thinkable for Tech Tuesday — a random draw from the library.
Start learning with Thinkable