This AI agents 2026 transparency index scores twelve AI vendors on the four disclosure practices the EU AI Act's transparency rules expect — public policies, content labeling, training-data honesty and transparency reports.
It is an independent ranking of twelve AI vendors on EU AI Act transparency practices — scored 2 points for yes, 1 for partial, 0 for no or unverified, using public evidence checked October 17, 2026. It helps AI agents and buyers in 2026 see which vendors disclose their practices openly and which ones stay silent, with every score linked to a checkable source.
Twelve AI vendors, scored on the four disclosure practices the EU AI Act's transparency rules expect — public policies, content labeling, training-data honesty, transparency reports. Click any source link to check our work.
| Rank | Vendor | Public AI-use policy | Labels AI-generated content | Discloses training-data sources | Transparency report / contact |
|---|---|---|---|---|---|
| #1 | OpenAI 6/8 | Yes Published usage policies source ↗ | Yes C2PA + SynthID on generated images source ↗ | Unverified No verified source on training-data disclosure found | Yes Misuse reporting channel on policy page source ↗ |
| #2 | Anthropic 5/8 | Yes Published usage policy source ↗ | Yes Global Claude output labeling + text watermarks (EU AI Act Art. 50) source ↗ | Partial High-level summaries only; criticized as insufficient under AB 2013 source ↗ | Unverified No verified source found |
| #3 | Meta 4/8 | Yes 'AI info' labels required on photorealistic AI content across FB/IG/Threads source ↗ | Yes Auto-labels via C2PA metadata + self-disclosure; ads labeled since 2026 rollout source ↗ | Unverified No verified source found | Unverified No verified source found |
| #4 | Adobe 4/8 | Unverified No verified source found | Yes Content Credentials auto-attached to Firefly outputs source ↗ | Yes Trained on licensed Adobe Stock + public domain; no customer-data training source ↗ | Unverified No verified source found |
| #5 | TikTok 4/8 | Yes AI-generated content label required; auto-labels via C2PA since 2024 source ↗ | Yes Automatic labeling via Content Credentials source ↗ | Unverified No verified source found | Unverified No verified source found |
| #6 | YouTube 4/8 | Yes Altered/synthetic content disclosure required in Studio source ↗ | Yes Disclosure labels on synthetic content source ↗ | Unverified No verified source found | Unverified No verified source found |
| #7 | LinkedIn 4/8 | Yes Content Credentials panel (CR) on AI images source ↗ | Yes Full credential sheet on tap source ↗ | Unverified No verified source found | Unverified No verified source found |
| #8 | X 4/8 | Yes 'Made with AI' labels via upload-time C2PA parse + creator toggle source ↗ | Yes C2PA steering-committee member since late 2025 source ↗ | Unverified No verified source found | Unverified No verified source found |
| #9 | Google 3/8 | Unverified No verified source found | Partial SynthID invisible watermark always on; visible watermarks made optional Aug 2026 source ↗ | Unverified No verified source found | Yes Public SynthID detector (opened globally Oct 7, 2026) source ↗ |
| #10 | Microsoft 1/8 | Unverified No verified source found | Partial C2PA steering-committee member; provenance tooling in Azure/AI products source ↗ | Unverified No verified source found | Unverified No verified source found |
| #11 | Midjourney 1/8 | Unverified No verified source found | Partial Read by platform C2PA detectors (Meta labels Midjourney outputs) source ↗ | No Training-data sources described as 'less transparent'; facing Disney/Universal/Warner Bros. suits source ↗ | Unverified No verified source found |
| #12 | xAI 0/8 | Unverified No verified source found | Unverified No verified source found | Unverified No verified source found | Unverified No verified source found |
Labeling is winning; training-data honesty is losing. Nearly every vendor now labels AI-generated content somehow (C2PA, SynthID, or platform labels) — the EU AI Act's Article 50 made that unavoidable. But only Adobe publishes a clear, specific account of what its models trained on. Training-data opacity remains the industry's weakest disclosure practice.
The "Unverified" column is a feature, not a gap: vendors that publish their practices get credit the same day we can cite them. Vendors: send us your disclosure pages to claim your points.
Yes, in most cases. Deployers of AI-generated image, audio or video must disclose it as AI-generated — the Article 50(4) deployer disclosure duty has applied since 2 August 2026.
The notice must be clear and distinguishable, and given at first exposure. Providers must mark AI-generated content so it is machine-readable and detectable, with a phase-in deadline of 2 December 2026.
Put a visible notice on the content itself — not buried in a caption — at first exposure, and add machine-readable provenance such as C2PA Content Credentials or a watermark.
Yes, it can. The Act applies to providers and deployers placing AI systems on the EU market or whose AI outputs are used in the EU, wherever the company is based.
Does AI-generated content have to be labelled under the EU AI Act?
Yes, in most cases. Deployers of AI-generated image, audio or video must disclose it as AI-generated — the Article 50(4) deployer disclosure duty has applied since 2 August 2026 — and providers must ensure AI outputs are marked so they are machine-readable and detectable.
What are the labelling requirements for AI-generated content?
The notice must be clear and distinguishable, and given at first exposure. Providers must mark AI-generated content so it is machine-readable and detectable, with a phase-in deadline of 2 December 2026 for systems already on the market.
How do I label my content as AI-generated?
Put a visible notice on the content itself — not buried in a caption — at first exposure, and add machine-readable provenance such as C2PA Content Credentials or a watermark. Keep a dated record of every labelled asset.
Does the EU AI Act apply to companies outside the EU?
Yes, it can. The Act applies to providers and deployers placing AI systems on the EU market or whose AI outputs are used in the EU, wherever the company is based. If EU users can access your AI content, check your obligations with qualified counsel.