Label the Real, Not the Fake: Instagram's Bet on Cryptographic Provenance
Last updated: January 1, 2026 7:03 pm
Instagram chief Adam Mosseri is pushing a simple idea with big impact: don't chase every deepfake after it spreads - label what's real at the point of capture. If a photo or video can prove where it came from and what happened to it along the way, people don't have to guess.
For PR and communications teams, this shifts the playbook from detection to proof. Authenticity becomes a feature, not a forensic project.
Why It Might Be Better to Label Reality Than Chase Fakes
Detectors are in an arms race. Watermarks get scrubbed, metadata gets stripped, and open-source models change faster than platform rules. You'll always be behind if your plan relies on catching every fake.
Signing authentic media at capture flips the script. Hardware or capture-app signatures use public-key cryptography to preserve a verifiable chain of custody through edits, exports, and posts.
The trust gap is real. In the Reuters Institute's Digital News Report, 59% of respondents say it's hard to tell real from fake online - even worse on visual platforms where content moves fast. Source
How Cryptographic Provenance Would Function
The standard gaining traction is C2PA (Coalition for Content Provenance and Authenticity). Cameras and capture apps sign media at creation and attach tamper-evident metadata describing when, where, and how it was captured. Viewers can see "Content Credentials" that summarize that history. Learn about C2PA
Camera makers are moving: Leica has shipped Content Credentials, Sony has field-tested in-camera signing with news orgs, and Nikon has announced support. Newsrooms like AP and Reuters are onboard so editors - and audiences - can verify a photo's journey pre-publication.
Software is catching up too. Adobe's tools can attach Content Credentials to workflows, and research groups are building more durable watermarks for synthetic output. The goal isn't to ban synthetic media; it's to give people a clear signal when something started in the physical world.
How Platform Policies Are Evolving on AI and Authenticity
Platforms are labeling AI-driven content and asking creators to disclose synthetic media. YouTube adds creator disclosures and labels for sensitive topics; TikTok requires labels for deepfakes; Meta labels AI content it detects or that creators flag.
The gap: detection is imperfect and disclosures are easy to skip. Mosseri's approach lines up with a bigger shift toward verifiable signals that survive edits and reposts as models get harder to spot. Regulators, especially in the EU, are nudging platforms toward deepfake labeling and standards-based provenance.
The Trade-Offs and Risks of Cryptographic Provenance Labels
- Pipeline fragility: authenticity breaks if any tool strips metadata. Video is especially tricky due to exports, compression, and reposting.
- Privacy: activists, journalists, and everyday users may not want location or device details exposed. Standards are working on selective disclosure, but UX must make the trade-offs obvious.
- Realness privilege: accounts with signed capture could get algorithmic preference, potentially disadvantaging creators without compatible devices.
- Context still matters: a real clip can mislead when cropped or miscaptioned. Mosseri has emphasized broader "credibility signals" - account history, affiliations, and past corrections - not just a green check on pixels.
What To Watch Next as Provenance Tech Rolls Out Broadly
- Firmware updates from major camera brands and pressure on smartphone platforms to make provenance a default capture option.
- Clear, consistent Content Credentials across Instagram, Facebook, and other apps - including for edited or remixed media.
- News orgs publishing more assets with verifiable histories, especially around elections and breaking events.
What PR and Communications Teams Should Do Now
- Audit your content pipeline: cameras, phones, editing tools, compression, scheduling, and publishing. Identify where metadata can break and fix it.
- Turn on Content Credentials in creative software and test end-to-end with sample assets. Document the exact export settings that preserve signatures.
- Talk to vendors: ask your camera, MDM, DAM, and social tools about C2PA support and retention of provenance data through transformations.
- Update disclosure policies: define when to label synthetic or heavily edited content, and how to display provenance to audiences.
- Prepare incident playbooks: when a deepfake targets your brand or executives, you'll need signed originals, rapid side-by-side comparisons, and a distribution plan.
- Align with legal and privacy: decide what metadata you'll show, what you'll redact, and how to explain it to stakeholders.
- Train spokespeople and social teams: teach them how to verify user-generated content, read Content Credentials, and communicate authenticity in plain language.
- Measure trust signals: track engagement and sentiment on posts with credentials vs. without; iterate your standards based on what builds confidence.
The Bottom Line
Misinformation won't be "solved" by better detection. Mosseri's point is pragmatic: if AI looks like reality, the internet needs receipts. Signing real media at capture lets platforms help people trust what they see without playing guesswork.
If you make content, label reality. If you manage brand risk, demand provenance. Both moves pay off the next time a fake runs faster than the truth.
Want practical upskilling for your team? Explore AI courses by job role to build skills in content authenticity, policy, and workflows: Complete AI Training - Courses by Job
Your membership also unlocks:
