No surprise: @clemwl.bsky.social from Google told policymakers a Pixel 10 photo is "real as if it’s just light hitting the lens".
In reality, Pixel images are explicitly tagged as computational photography: AI‑reconstructed.
"Computational Capture" was introduced by IPTC after Google asked them.
Nice to see that @adobe.com @truepic.bsky.social @andyparsons.net and @mouniribrahim.bsky.social lobbying never getting old despise aging like milk.
The video is almost 2 years old. no change so far.
bsky.app/profile/bots...
Meet Leonard Rosenthol: C2PA Architect, PDF Architect, ISO TC171 SC2 Chairman, Adobe Senior Principal Scientist and member of everything else.
And yes, he excavated a 34 year old obscure spec to bury C2PA in it, breaking the internet, and will lecture anyone brave enough to ask why.
Even if better than the deeply flawed "nutrition label", the analogy "C2PA = TLS for media" still doesn’t hold. The two systems solve different problems, operate at different layers (application vs network), and protect different things (content vs connection).
A lot of questionable claims have been made about C2PA, particularly from AI‑driven companies like Adobe and Microsoft (OpenAI), with intense PR and lobbying by the astroturfing group @contentauth.bsky.social. It raises the concern that the motivation may have been to avoid potential AI regulation.
Vivo has a Camera, if you trust the C2PA conformance. 😉
bsky.app/profile/bots...
Hello Nathan!
"C2PA=TLS for media" was for last year marketing material.
You should now use: C2PA=App-signing for media, per Google @sherifhanna.bsky.social
2020 → Authenticity
2021 → Attribution
2022 → Provenance
2023 → Nutrition label
2024 → Bottle seal
2025 → HTTPS
🆕 2026 → App-signing
The Verge @seanhollister.bsky.social worries that "[...]rhetoric, and the C2PA itself, are substitutes for actual meaningful action."
www.theverge.com/tech/885727/...
Don't worry, @contentauth.bsky.social has been lobbying all along for the AI industry: C2PA never worked.
bsky.app/profile/bots...
Yes! It got worse!
@adobe.com @contentauth.bsky.social is promoting a genAI advertorial they probably paid for, on the 100% AI Slop Social Network Linkedin, published on a AI content farm bloated with ads for AI Slop. 🤦♀️
The AI Content Farm link:
www.techtimes.com/articles/314...
This @adobe.com proof of concept has been discontinued and is no longer supported. Content credentials can be removed easily, offering no real protection for artists.
What is worse?
1. @adobe.com advertising on AI Slop farms ?
2. @contentauth.bsky.social probably buying advertorials for #C2PA on the same AI Slop farms ?
Note: The headshot is from the AI model that generated this article.
www.diyphotography.net/content-cred...
You can skip it.
Prompt: A banana peel floating over a blurry CRT displaying a bloated C2PA manifest abusing Unicode.
A disappointing step backward for metadata standards.
Google was the last major platform still preserving IPTC for its genAI Slop, a format readable by nearly any tool from the past two decades.
Now it embeds only C2PA: a bulky, poorly supported format that no software can decode, except Google.
Fun fact for anyone who prefers cannibalism over dogfooding for their AI slop needs (part 2):
One of the video promoting "Gemini 3 Deep Think", as a peer review peer, includes a 00:00:13:06=>00:00:16:06 insert of a video generated by the genAI model kling 3.0.
blog.google/innovation-a...
It only took two whole months before someone finally create an issue that Google C2PA implementations are basically DoS‑ing the official @adobe.com @contentauth.bsky.social validators.
github.com/contentauth/...
"The bad news is that Instagram is already using [C2PA], and it’s not doing shit to actually help. If anything"
Another banger by @zombiewretch.bsky.social with @andyparsons.net still promising, after 5 years, that "it's coming..." while most POC have been trashed.
www.theverge.com/ai-artificia...
It appears that Adobe’s C2PA is now working with the Information Technology Industry Council (ITI) lobby group rather than @contentauth.bsky.social
Why C2PA was in India ? : Adobe’s agreement with the government to provide "free" "AI‑first learning" to students.
www.gadgets360.com/ai/news/ai-i...
AI content farms already strip C2PA automatically. The available verification code is non‑compliant, immature, and computationally heavy.
Social-Media have no interest in C2PA, and no independent research demonstrates its effectiveness beyond Adobe’s claims.
C2PA isn’t mature or interoperable, and no compliant verifier exists after 5 years. Its AI part is just vaporware. By 2026, Adobe seems to have moved on, except when government money is on the table: Don't rely on a deceptive company.
URL fixed: drive.google.com/file/d/1VMkN...
Meet Paul Melcher, "Expert in visual content authenticity and provenance", member and evangelist of the #C2PA.
He has just launch a new #C2PA "Translator in Plain english" verifier.
Of course I had to try it with some random forgery. 🤡
Content Credentials (C2PA) are now v2 and must be compliant.
Adobe still run deprecated 5 years old v1, in beta, in most of its software (Lightroom)
It is not compatible, or interoperable as Adobe claims. It's just a marketing claim to lock-in users with subscriptions.
bsky.app/profile/bots...
Meet Erik Svilich: co‑chair of the #C2PA Text Task Force and the creator of the Unicode‑based hack that enabled embedding C2PA metadata directly into plain text.
It somehow made its way into the official C2PA specification: without consulting Unicode.
His previous venture: AI Avatars Slop.
Surprise! Ce qu'il reste de @contentauth.bsky.social ( Andrew Jenks & @andyparsons.net ) partent en Inde pour un dernier tour d'honneur de lobbying: refourguer de l'IA @adobe.com &Co. "éthique" aux écoles et administrations, avec le soutien du BSA, sous couvert de #C2PA.
www.bsa.org/files/policy...
It's obviously an @adobe.com #abandonware, except some vibe coding by Google @sherifhanna.bsky.social
No news from @contentauth.bsky.social either.
They may be busy lobbying in India, another round for regulatory capture?
@https://bsky.app/profile/botsdontcry1.bsky.social/post/3mbhgokcpss2k
❤️ AI?
❤️ facial recognition?
❤️ computational photography?
❤️ C2PA?
Combine all your favorite hobbies with the C2PA-compliant Xiaomi Redmi Note 15 5G!
A device probably designed to slop out "authentic" AI‑generated images using its Megvii AI model.
Bonus: it's trained by mass-surveillance.
Albert Uderzo: Astérix le Gaulois
Adobe tried to dismiss a patent lawsuit against C2PA, brought by French software company Keeex, by claiming that an evidence was only a digital copy, not the original.
Keeex could have used C2PA to authenticate the document.
If it actually worked.
www.unifiedpatentcourt.org/sites/defaul...
Verify first!
C2PA isn’t a standard, its ISO draft was rejected, and remains just an Adobe defined specification.
Its implementations aren’t interoperable. Lightroom’s version embeds proprietary Adobe metadata that existing readers can’t interpret, and it’s already deprecated and non‑conformant.
C2PA has been a success.
Its flaws were obvious when the AI industry sold it to governments as a replacement for real regulation.
They knew it was not working and lied about it.
They avoided regulation. C2PA likely helped.
And that's how you make money with C2PA.
www.theverge.com/podcast/8740...
C2PA is just a glorified image signer, it can't tell if an Image is AI generated or not. At best it gives information whether you can trust the issuer.
verify.contentauthenticity.org is a 5 years old Adobe demo, deprecated with a misleading UI. It should not be used.
bsky.app/profile/bots...
Adobe is so desperate! It is joining the AI cash burning train by subsidizing its entire genAI models wrapper (Firefly).
Poor Adobe Stock 🤣
blog.adobe.com/en/publish/2...
If you truly think you need genAI (you don’t), there are plenty of better subsidized wrappers slopping around.
As of February 1, Google has enabled SynthID (invisible watermark layer for genAI) across all Gemini plans/APIs/partners. C2PA remains, with a proprietary tag that Adobe’s verifier can’t surface.
Anyone running serious botfarms pipelines knows it’s still removable, at the cost of extra computation.
