Spain's AI bill and AI labels: in Congress, not law yet
Spain's AI governance bill reached Congress in June 2026 with fines up to 35 million euros. Is AI video labelling Spanish law yet, and what to check on a file?

No. Spain's Organic Law Bill for the good use and governance of artificial intelligence is still in parliament. Press reports say the Council of Ministers approved it on 26 May 2026, and a law-firm alert says it began its passage through Congress on 12 June 2026, with amendments accepted until 30 June. Until it passes, the labelling duty that applies to AI video in Spain is the EU's own, in Article 50 of the AI Act.
These facts come from news and commentary, not the bill's text, which I did not read: Finreg360, SapientiaDEV and Economist & Jurist, all read on 2026-10-03. It is not legal advice.
What does the bill add to the EU rule?
SapientiaDEV describes it as adding enforcement machinery and national penalties rather than new material requirements, and says the EU regulation's obligations already apply now. Economist & Jurist says the bill covers identification of AI-generated content, notice to people who interact with automated systems and marking of synthetic outputs, and gives general penalty tiers rather than a separate sanction for a labelling failure. For when the EU duty bites on older systems, see the 2 December 2026 grace period.
| Tier | Maximum fine as reported |
|---|---|
| Prohibited practices | 35 million euros or 7% of global turnover |
| Other very serious violations | 15 million euros or 3% |
| Serious infractions | 7.5 million euros or 1% |
| Minor infractions | 500,000 euros or 0.5% |
What about the 2025 draft with the 35 million euro fines?
That was an earlier text. Euronews reported on 12 March 2025 that the cabinet had adopted a draft on 11 March 2025, fining incorrect labelling of AI content between 7.5 and 35 million euros, or 2 to 7% of global turnover, with lighter penalties possible for startups and medium-sized firms. The tiers in the table above are from the 2026 bill; do not mix the two when you quote a figure.
What should I check on the file before it ships to Spain?
Marking, if you rely on it, is a property of the file you deliver, not of the tool that made it. Sume's docs do not say a video output carries a C2PA record or watermark, which an earlier post covers, so test the file you ship rather than assume it. Sume's video trim shows why the file changes: precision: "exact" is a frame-accurate re-encode (libx264, yuv420p), while keyframe is a stream copy whose cut may start a GOP early. Each trim returns a new MP4 and leaves the source untouched.
So run your marking check on the output of the last step, not on the first render. Content Credentials after download shows how, and the transparency Code of Practice versus the Article 50 guidelines sorts which text asks what. For a reviewer's read of the finished cut, video inspect returns stills and an optional transcript without producing a new file.
Sources
- Euronews: Spain could fine AI companies up to 35 million euros for mislabelling content (read 2026-10-03)
- Finreg360: El Congreso inicia el trámite del proyecto de ley orgánica para el buen uso y la gobernanza de la IA (read 2026-10-03)
- SapientiaDEV: Ley española de IA, qué se aprobó de verdad el 26 de mayo (read 2026-10-03)
- Economist & Jurist: Ley Orgánica de IA, España aterriza el AI Act (read 2026-10-03)
- Video trim
- Video inspect
Related posts
More in Use cases
- Spanish song lyrics from an AI music prompt: write it in Spanish
Google says Lyria lyrics follow the prompt language. For a Spanish track on Sume, write the prompt and lyrics in Spanish, then read the lyrics section map.
- A speaker withdrew voice consent: finding every narration that used it
Revocation is a clean-up job. Build a manifest of voice, model and job id for each narration, then regenerate with another voice and join audio on Sume.
- Splice an AI ending onto a Short and keep the original audio
Gemini Omni clips come with their own sound. To keep your Short's voice and music, detach both audio tracks and join them as audio parts in one Timeline render.
- Split a Short's audio at a cut point: timeline-audio ranges
YouTube Create has a split tool for audio. Sume's timeline-audio endpoint splits one file into up to 20 ranges, or joins up to 20 parts, for $0.01 a job.
Written by Sume