Google opened SynthID.com to anyone, checking images, video and audio for OpenAI, Nvidia and Kakao watermarks as well as its own. That speeds up flagging partner-made media but cannot vouch for files from outside the SynthID group.
Perspective Coverage
20 publishers
- Builder
- Builder 27%
- Operator
- Operator 54%
- Investor
- Investor 19%
Reality
- Evidence72
- Adoption40
- Hype gap+15
- Incentives55
- Confidence68
buildOne report1 publisher OpenAI's textGrain text detector goes to approved researchers first
OpenAI is giving its textGrain watermark detector for ChatGPT and Codex text only to approved researchers and expert organizations, case by case. Teams reviewing outside documents have no public checker and no date for one, so their AI-content policies have to rest on their own records.
Reality
- Evidence40
- Adoption
- Not scored
- Hype gap+10
- Incentives55
- Confidence35
buildOne report1 publisher Anthropic embeds a watermark in Claude's wording that survives copying into other apps
Anthropic now embeds a token-free watermark in Claude output across its API, Claude Code and three cloud partners. For content teams that publish Claude drafts, the practical work now sits in disclosure policy and internal records.
Reality
- Evidence38
- Adoption45
- Hype gap+12
- Incentives62
- Confidence40
buildOne report1 publisher Google asks publishers to hand-check AI-written titles, schema and alt text before release
Google's updated AI-content guidance calls manual fact-checking critical for titles, meta descriptions, structured data and alt text as well as article text. Its firm requirements are reserved for ecommerce provenance labels, so for most publishers human review is strong advice that a pipeline feeding model output straight to a CMS skips at a quality cost.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+20
- Incentives70
- Confidence45
buildOne report1 publisher UCLA's optical deepfake screen classifies at least 15 videos in one pass of light
UCLA researchers' hybrid optical-neural processor screens at least 15 video streams per pass of light at 97.79% accuracy on Celeb-DF. The experimental eLight study pitches it as a cheap first filter whose savings hinge on how many genuine videos reach the heavier detectors.
Reality
- Evidence55
- Adoption3
- Hype gap+10
- Incentives35
- Confidence50
OpenAI leads on lighting and speed, but the claims that decide whether a team can ship an asset are about holding a subject consistent across turns and leaving untouched regions alone, and neither arrives with a number.
Perspective Coverage
7 publishers
- Builder
- Builder 22%
- Operator
- Operator 64%
- Investor
- Investor 14%
Reality
- Evidence45
- Adoption30
- Hype gap+30
- Incentives55
- Confidence60
buildOne report1 publisher Article 50 makes the export pipeline carry the AI provenance
The EU's synthetic-content transparency duties took effect on 2 August 2026, and they ask for two different artifacts: a machine-readable mark that survives every export, and a human-readable notice in every channel the asset lands in.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+15
- Incentives70
- Confidence45
buildOne report1 publisher Lootr Tools refuses a drop rate that arrives without a provenance status
An unofficial Dungeon Lootr companion runs every class, item and build through a Zod schema at import, then cross-checks the whole dataset. A guide pointing at a class that was never defined fails at build.
Reality
- Evidence45
- Adoption12
- Hype gap+10
- Incentives40
- Confidence60
macOS 27 folds three generative image tools into the everyday Photos editor
Three tools in the new Photos editing pane synthesize image data inside ordinary consumer libraries. Shared Albums in the same release move those files at full resolution, including to Android and Windows contributors.
Reality
- Evidence42
- Adoption34
- Hype gap+18
- Incentives68
- Confidence46
Apple signs iPhone photos inside the camera hardware before software touches the pixels
Reference Image on the iPhone 18 Pro puts one signature on the sensor's raw output and a second on the developed file. The capture time it proves is only good to about 15 minutes.
Reality
- Evidence48
- Adoption8
- Hype gap+20
- Incentives62
- Confidence55
buildOne report1 publisher A sha256 sidecar now decides what the posting agent is allowed to attach
After counting 42 scheduled Bluesky posts in a week with five images, one team cut its poster's reach to a single committed directory and made the commit gate and the poster read media through the same hashing loader.
Reality
- Evidence52
- Adoption12
- Hype gap+20
- Incentives55
- Confidence45
Decrypt's rerun gives three of six categories to Google and the rest to OpenAI. The 50% latency cut is OpenAI's own figure, measured against Images 2.0, and the review does not include prices for the two new API tiers.
Reality
- Evidence42
- Adoption45
- Hype gap+38
- Incentives58
- Confidence50
Google shipped a prompt box into the satellite tool investigators are trained on
The generative feature inside Google Earth lasted about a day before the outcry pulled it, but the verification checklists that name Google Earth as ground truth were written for a product nobody could prompt.
Reality
- Evidence56
- Adoption28
- Hype gap+18
- Incentives62
- Confidence52
Substack hands its readers a button that scans your post for AI text
Fast Company reports that Pangram beats rival detectors in independent testing, without naming an error rate, and Substack, NewsGuard and an exam platform are already running it on writing whose authors never opted in.
Reality
- Evidence42
- Adoption58
- Hype gap+32
- Incentives66
- Confidence48
Adobe wired OpenAI's Images 2.5 into Firefly on launch day
OpenAI's newest image models now render inside Adobe's creative studio while more than 70 Adobe tools sit inside ChatGPT, which leaves each company distributing the other and neither owning the moment a creative task starts.
Reality
- Evidence34
- Adoption55
- Hype gap+14
- Incentives74
- Confidence46
buildOne report1 publisher Re-encoding the pixels is the step that actually drops a C2PA manifest
Generated images arrive with three metadata layers plus a pixel signal that no metadata tool touches, and whether the signed manifest survives depends on whether anything in your pipeline decodes and writes the file again.
Reality
- Evidence60
- Adoption
- Not scored
- Hype gap+12
- Incentives70
- Confidence48
buildOne report1 publisher C2PA proves who signed, not what was true: Content Credentials are provenance, not AI detection
The standard signs a chain of assertions about capture and edits. It does not verify the scene in front of the lens, and a missing manifest tells you almost nothing.
Reality
- Evidence42
- Adoption
- Not scored
- Hype gap−12
- Incentives22
- Confidence48