EU AI Act: transparency obligations for businesses from August 2026

European businesses face a paradigm shift. AI adoption in operational processes is growing at an unprecedented pace: according to McKinsey’s State of AI report, 88% of organizations already use AI in at least one business function. In Italy alone, ISTAT data shows AI adoption among businesses doubled from 8.2% in 2024 to 16.4% in 2025.

Yet since August 2, 2026, Regulation (EU) 2024/1689, known as the EU AI Act, has made stringent transparency requirements applicable to anyone developing or deploying certain AI systems, such as chatbots and content generators; for AI agents, governance also starts with certifying AI agent data. Article 50 imposes specific obligations around informing users, machine-readable marking and labeling of AI-generated content. The problem: many businesses are still not ready.

The answer is not last-minute compliance patches. It lies in building a Digital Provenance infrastructure that certifies the origin and authenticity of digital content at the source, turning compliance from a reactive burden into a structural competitive advantage.

EU AI Act: the first comprehensive AI regulation

The EU AI Act is the world’s first comprehensive regulatory framework for artificial intelligence. Entered into force on August 1, 2024, it introduces obligations progressively over several years.

Key milestones:

  • February 2025: AI literacy (Art. 4) and prohibition of unacceptable-risk practices (Art. 5) already in force
  • August 2025: obligations for general-purpose AI model (GPAI) providers, designation of national competent authorities, penalty regime activated with fines up to EUR 35 million or 7% of global turnover
  • August 2, 2026: transparency obligations (Art. 50) apply. Providers of systems generating synthetic content placed on the market before that date have until December 2, 2026 to comply with the machine-readable marking required by Art. 50(2)
  • December 2, 2027 and August 2, 2028: requirements for high-risk AI systems, postponed by Regulation (EU) 2026/1744 (Digital Omnibus) for Annex III and Annex I systems respectively

The enforcement structure relies on national competent authorities in each Member State, coordinated through the European AI Board. Penalties apply to both EU and non-EU companies offering AI systems in the European market, making this a truly extraterritorial regulation with teeth.

Article 50: transparency requirements in detail

Article 50 of the AI Act introduces differentiated obligations for providers and deployers of AI systems, with particular focus on AI-generated content. What each of them has to do, and the grace period that runs until 2 December 2026 for the machine-readable marking of systems already on the market, is covered in our insight on Article 50 and the labelling of synthetic content.

Obligations for AI system providers

Developers of AI systems must ensure three fundamental requirements:

  • AI systems interacting directly with people must inform users they are communicating with an artificial intelligence, unless obvious from the context
  • Generated content (text, audio, images, video) must be marked in a machine-readable format and detectable as artificially generated
  • Marking solutions must be effective, interoperable, robust and reliable, as far as technically feasible

Obligations for deployers

Organizations deploying AI systems in their processes carry equally specific responsibilities:

  • Disclose that deepfake content has been artificially generated or manipulated
  • Label AI-generated or manipulated text published to inform the public on matters of public interest, unless it has undergone human review or editorial control and a person holds editorial responsibility for it
  • Provide this information in a clear and distinguishable manner at the latest at the time of the first interaction or exposure, in line with applicable accessibility requirements
  • Related measures that do not stem from Art. 50: AI literacy of staff using AI systems (Art. 4) and human oversight procedures

The Transparency Code of Practice

Alongside Article 50, the European Commission has promoted a Code of Practice on transparency of AI-generated content, final since June 10, 2026, and on July 20, 2026 it published its guidelines on the transparency obligations. The first draft, released in December 2025, already outlined a multi-layered approach: metadata embedding with provenance standards (including C2PA), imperceptible watermarking at the pixel level, and logging systems where other techniques prove insufficient. The Code is voluntary: the Commission and the AI Board have assessed it as an adequate tool, and by the end of July 2026 around 190 organizations had signed it.

The readiness gap across European businesses

The data on European business readiness reveals a concerning gap. A Deloitte survey of 500 managers found that only 35.7% feel adequately prepared for AI Act compliance, while 19.4% describe themselves as poorly prepared. Just 26.2% have actually started concrete compliance activities.

Structural gaps are widespread:

  • Over 50% of organizations lack a basic inventory of their AI systems
  • Only 18% have implemented a complete AI governance framework
  • 70% have no formalized AI governance model
  • Only 28% have CEO-level oversight and 17% have board-level oversight of AI

The penalty regime, already active for prohibited practices, provides for fines of up to EUR 35 million or 7% of global annual turnover for the most serious violations, up to EUR 15 million or 3% for other non-compliance (including transparency obligations), and up to EUR 7.5 million or 1% for providing inaccurate information to authorities.

The challenge for small and mid-size enterprises

While large corporations are accelerating AI adoption (53.1% in Italy, according to ISTAT), SMEs lag significantly behind at 15.7%. The gap widened by 12 percentage points in a single year. For SMEs, the dual challenge of limited AI expertise and compliance complexity creates a particularly difficult environment. The 60% of companies that evaluated AI investments without proceeding cite lack of adequate skills as the primary barrier.

Data provenance: the structural approach to compliance

Complying with the AI Act is not about adding labels to AI-generated content after the fact. It requires building an infrastructure capable of tracing, verifying, and certifying the origin of every piece of digital content throughout its entire lifecycle.

This approach, grounded in Digital Provenance, follows a clear logic: in a world where any content can potentially be generated or manipulated by AI, trying to detect what is fake becomes increasingly futile. What works is guaranteeing what is authentic at the source, making every piece of data immediately verifiable and trustworthy.

How it works in practice

TrueScreen operates exactly in this direction. Its platform enables organizations to acquire, verify, and certify any digital content (photos, videos, documents, emails, web pages) with legal and probative value.

The process is built on a patented forensic methodology that follows ISO/IEC 27037, and TrueScreen is ISO/IEC 27001 certified. Upon acquisition, the certification integrates a qualified electronic seal and a qualified timestamp under the eIDAS Regulation (EU 910/2014), recognized throughout the European Union. Cryptographic hashing algorithms ensure that any subsequent modification is detectable.

The certified output is a package that includes the XML file sealed with a qualified electronic seal and a qualified timestamp, the report data in JSON, the PDF report and, when the acquisition has media, the original files. This package creates a verifiable audit trail of the origin and integrity of authentic content: it supports the traceability and provenance obligations of the AI Act, while marking AI-generated output remains a duty of the providers and deployers of AI systems.

From reactive compliance to proactive infrastructure

The fundamental difference between surface-level compliance and a structural strategy lies in sustainability. Adding watermarks or metadata on a case-by-case basis is fragile, generating rising costs and coverage gaps. Integrating source certification into existing operational workflows through the TrueScreen API makes the compliance process more automated and scalable.

TrueScreen follows ISO/IEC 27037 and integrates a qualified electronic seal and a qualified timestamp under eIDAS: with certified data it provides evidence of origin and integrity that organizations can use to document their own compliance, for example under the AI Act and NIS2. Compliance remains with the organization that adopts it.

Operational roadmap: meeting Article 50 obligations

Transparency obligations have applied since August 2, 2026, and systems already on the market before that date must complete the Art. 50(2) marking by December 2, 2026. Here are the five concrete steps every organization should take, if it has not already.

1. Map all AI systems in use: create a comprehensive inventory of AI systems deployed across the organization. For each system, identify: the provider, the type of output generated (text, images, audio, video), the risk level, and the business processes involved. According to the surveys above, over 50% of European organizations had not yet completed this fundamental first step.

2. Classify by risk level: the AI Act distinguishes four risk levels: unacceptable (prohibited), high, limited, and minimal. High-risk systems (Annex III) require technical documentation, automatic logging, and human oversight, with requirements applying from December 2, 2027. Article 50 obligations concern specific types of systems (those that interact with people, generate synthetic content or recognize emotions) and also apply to systems classified as high-risk. Correctly classifying systems is the prerequisite for defining the required compliance actions.

3. Implement content traceability: for authentic content the organization acquires, certify origin and integrity with verifiable metadata, a qualified electronic seal and a qualified timestamp; for AI-generated content, check that the systems in use apply the marking required by Art. 50(2) and keep the related documentation. Integrate these solutions into existing workflows via APIs to make the process more automated and scalable.

4. Build internal governance and training: since February 2025 the AI Act has addressed the AI literacy of staff (Art. 4). Define clear roles and responsibilities, implement internal policies on AI usage, train teams on managing AI-generated content, and establish human oversight mechanisms.

5. Monitor regulatory developments continuously: the regulatory landscape continues to evolve. The Transparency Code of Practice has been final since June 2026, and Regulation (EU) 2026/1744 (Digital Omnibus), in force since July 27, 2026, has already moved some deadlines, starting with those for high-risk systems. A proactive approach requires ongoing monitoring and timely process adaptation.

FAQ: EU AI Act transparency obligations for businesses

When do the EU AI Act transparency obligations take effect?
The transparency obligations under Article 50 of the EU AI Act apply from August 2, 2026. For systems generating synthetic content placed on the market before that date, the Art. 50(2) marking obligation must be met by December 2, 2026, under Regulation (EU) 2026/1744. Some related obligations were already in force: rules on AI literacy (Art. 4) have applied since February 2025, and obligations for general-purpose AI models since August 2025.
Which businesses are subject to the AI Act transparency requirements?
The obligations apply to both providers (who develop AI systems) and deployers (who use them in their processes) operating in the European market, regardless of where the company is headquartered. For SMEs, including start-ups, fines follow the lower of the fixed amount and the percentage (Art. 99(6)), but they are not exempt from the substantive obligations.
What does data provenance mean in the context of the AI Act?
Data provenance is the ability to document the origin, history, and transformations of digital content throughout its entire lifecycle. In the AI Act context, it is the basis for documenting the origin of content: for authentic content, by certifying its acquisition and integrity; for AI-generated content, through the marking and metadata that system providers must apply (Art. 50(2)).
What penalties does the EU AI Act impose for non-compliance?
The penalty regime has three tiers: up to EUR 35 million or 7% of global annual turnover for prohibited practice violations; up to EUR 15 million or 3% for other non-compliance (including transparency obligation violations); and up to EUR 7.5 million or 1% for providing inaccurate information to authorities.
How can a business start preparing for the transparency obligations?
The fundamental steps are: map all AI systems in use, classify them by risk level, implement data provenance solutions to document authentic content, train staff, and define internal governance. Integrating source certification into operational workflows via APIs makes documenting the origin and integrity of data automated and scalable; compliance remains with the organization.

Protect the authenticity of your digital content

The EU AI Act sets new transparency standards. TrueScreen certifies the authenticity of your data at the source and documents its origin, supporting the traceability and provenance obligations of the AI Act.

mockup app
TS

TrueScreen editorial team

Published on · Updated on

This section is edited by the TrueScreen editorial team, which brings together expertise in digital forensics, the law of digital evidence and regulatory compliance. Every article is checked against primary sources: legislation, published rulings, technical standards and official documentation, always cited in the text.

Editorial Team and Methodology →