EU AI ACT ART. 50 APPLICABLE 2 AUG 2026 · MACHINE-READABLE MARKING ENFORCEABLE 2 DEC 2026 ART. 50 TRANSPARENCY BREACHES — FINES UP TO €15M / 3% GLOBAL TURNOVER FINAL CODE OF PRACTICE (10 JUN 2026): SIGNED METADATA + WATERMARK + PUBLIC DETECTION EU AI ACT ART. 50 APPLICABLE 2 AUG 2026 · MACHINE-READABLE MARKING ENFORCEABLE 2 DEC 2026 ART. 50 TRANSPARENCY BREACHES — FINES UP TO €15M / 3% GLOBAL TURNOVER FINAL CODE OF PRACTICE (10 JUN 2026): SIGNED METADATA + WATERMARK + PUBLIC DETECTION
Blog / News

Insights on AI compliance
and the agentic economy.

Technical analysis, regulatory updates, and perspectives on building trust infrastructure for AI-generated content.


Recent articles.

AUG
2026
Regulation

Article 50 is live: what the final Guidelines and the adequacy-assessed Code actually change

The transparency obligations entered into application on 2 August. ~190 organisations signed the Code — and the regulator's own framing now describes a layered architecture: signed metadata, robust watermarking, and a registry for recovery.

JUL
2026
Regulation

The final Article 50 guidelines, and what actually comes due on 2 August

The Commission adopted the final transparency guidelines on 20 July. Google, OpenAI and Mistral signed the Code; Meta declined. Disclosure is due 2 August — and the guidelines concede the marking standard is still being developed.

JUL
2026
Regulation

The Transparency Code has been endorsed. What that actually changes before 2 August.

The Commission and the AI Board declared the Code adequate on 8–9 July. It is now the EU-wide instrument for Article 50 compliance — but signing is the beginning of the evidentiary conversation, not its conclusion.

JUL
2026
Regulation

Executive briefing: the final Transparency Code and what comes due on 2 August

Three dates now structure the work: 22 July (initial-signatory deadline), 2 August (Article 50 applies), 2 December (marking for pre-existing systems). And why signing the Code is not the same as complying.

JUN
2026
Regulation

The EU just described the architecture of Article 50 compliance

The 10 June Code of Practice narrows the technical question to a layered, signed-plus-watermarked-plus-verifiable model. The design problem is solved; the engineering problem is not — and the signatory deadline is 22 July.

JUN
2026
Regulation

The EU's Transparency Code is final. Here is what it actually changes.

On 10 June 2026 the European AI Office published the final Code of Practice on Transparency of AI-Generated Content. Voluntary in form, it is set to become the de facto compliance benchmark before Article 50 applies.

MAY
2026
Regulation

The Article 50 architecture has converged

Two regulatory events in 48 hours clarified both the Article 50 timeline and the expected technical architecture. The deadline provisionally moves to 2 December 2026, but the multilayer requirement remains.

APR
2026
Regulation

Invisible watermarking: why the AI Act requires it

Metadata alone is not enough. Under Article 50, AI-generated content must remain detectable even after compression, resizing, or platform re-uploads. Watermarking is the foundation layer.

APR
2026
Developer

The AI Act 50 Skill: making any AI agent a proactive Article 50 compliance agent

The MCP Server gives Claude the tools to certify AI-generated content. A SKILL.md gives it the judgment to know when certification is required — without human intervention.

APR
2026
Regulation

EU AI Act Article 50: What It Actually Requires (Definitive Guide 2026)

Article 50 introduces a clear obligation: all AI-generated content must be identified as such. But labeling is not enough — the regulator requires verifiable, multilayer compliance.

APR
2026
Regulation

C2PA is not enough: what the AI Office Code of Practice actually says

The Code of Practice expects a multi-layer system—not one isolated technology. Why C2PA metadata alone does not satisfy the EU AI Act.

MAR
2026
Blue

Why C2PA was designed for humans with tools — and what breaks when agents take over

C2PA requires an X.509 certificate, a compatible creation app, and a metadata-preserving channel. In the agentic economy, all three assumptions fail simultaneously.

FEB
2026
Blue

x402: how machine-to-machine HTTP payments enable autonomous AI compliance

Stripe is investing $1.1B in stablecoin infrastructure. AWS has shipped x402 support (Bedrock AgentCore Payments). 35M x402 transactions on Solana in March 2026. Here is why this matters for AI Act compliance.

ENE
2026
Green

The agentic layer on top of C2PA: how AI Act 50 extends the standard without replacing it

Adobe, Google, and Microsoft have invested years in C2PA. We don't compete with that — we complete it. A technical explanation of how AI Act 50 fits into the standards ecosystem.

DIC
2025
Amber

The €15M fine nobody talks about: who Article 50 actually catches and how

Most discussions focus on high-risk AI systems. Article 50's transparency obligations reach far more organisations — including deployers who assume marking is their model vendor's problem.

Technical analysis.
No noise.

Regulatory updates and protocol developments as they happen.