YouTube 'Info from' under an AI label: what the signing authority is
The 'Info from' line under a YouTube AI disclosure names who signed the Content Credentials. Which labels carry it, and what a Sume export does not add.

The 'Info from' line under a YouTube disclosure names the signing authority: the party that validated the information inside the video's secure Content Credentials (C2PA). YouTube's help page says Content Credential disclosures often include that attribution, and when they do, you see 'Info from' beneath the disclosure. It does not appear on a label that came only from your own answer in YouTube Studio.
That distinction decides what you can do as a creator. If you generate a clip with a tool that signs its output, the signer's name can travel with the file. If you export from a tool that does not sign, YouTube has no signer to name, so the label comes from your 'AI use' answer or from YouTube's own detection.
Where the label comes from
The YouTube help page on 'How this content was made' lists three routes into a 'Made with AI' disclosure. The table keeps them apart, because only one of them can carry a signer.
| Route | What triggers it | Can show 'Info from' |
|---|---|---|
| You disclose | You answer the AI question in YouTube Studio | No signer is involved |
| YouTube's own tools | You use YouTube generative tools such as DreamScreen | Not stated on the page |
| Content Credentials | Valid C2PA 2.1 or higher data says the entire video is made with AI | Yes, when the credentials name a signing authority |
What this means for a Sume export
Sume's video generation and trim docs describe the files and the job fields. They do not describe embedding or preserving Content Credentials, so do not plan on an 'Info from' line for a file that came from Sume. Plan on the manual route: answer the AI question yourself when the clip is realistic.
Trim is also a re-encode by default. The trim docs say precision: exact is a frame-accurate re-encode with libx264, and the result is a new artifact, never the source. A new file is a new file; any provenance data on the old one is not something the docs promise to carry over.
Steps
- Decide per clip whether it is realistic AI footage, using the examples on the GenAI disclosure page.
- In YouTube Studio, open the Attributes section and set 'AI use' to Yes or No.
- After upload, open the video and read the expanded description's 'How this content was made' section to see which label YouTube applied.
- If you see 'Info from', the label came from signed Content Credentials, not from your answer. Treat it as fixed.
- Record the clip's job id and your answer in your own log so the choice is auditable later.
Caveats
The help page says YouTube may also use Content Credentials to carry forward other disclosures about how content was made, so the section can hold more than 'Made with AI'. It also notes that YouTube can proactively apply a label to undisclosed content that creators will not have the option to remove. Missing 'Info from' therefore does not mean unlabeled, and an unwanted label may not be removable if it came from credentials or review.
Sources
Related posts
More in Developers
- YouTube Live 1080p30 H.264 bitrate: 5 to 14 Mbps, GB per hour
YouTube Live recommends 5 to 14 Mbps for 1080p30 H.264, with 2 B-frames and a 2-second keyframe. The gigabytes per hour for each rung, and what Sume cannot set.
- YouTube Shorts Creator Pool example: $405 from 1 million engaged views
YouTube's own worked example pays $405 for 1 million engaged views in a 100 million market. The math in Python, with the all-music case at $225.
- Turn a 16-shot list into 16 Sume video jobs in Python
Submit one /v1/videos job per shot with a first frame, run them concurrently with asyncio, and give each shot an Idempotency-Key so a rerun never bills twice.
- 402 halfway through 50 clips of 30-second video: stop and resume
When balance runs out mid-batch Sume answers 402 insufficient_credits before provider work starts. Halt, keep the paid job ids, and resume the rest later.
Written by Sume