Copyright Lawyers Greece 2026: Law 5179/2025, Ai‑training Permissions & Administrative Fines

Last updated: 8 May 2026

Greece’s copyright landscape has shifted decisively for any organisation that trains artificial‑intelligence models on works created, managed or archived within the country. Copyright lawyers Greece‑wide are fielding an unprecedented volume of compliance enquiries following the enactment of Law 5179/2025 and its successor, Law No. 5271/2026, which together impose stricter permission requirements for AI training data, significantly higher administrative fines for online piracy and unauthorised dataset use, and new rules governing cultural‑heritage materials held by museums and public institutions. This pillar guide explains every obligation that in‑house counsel, AI product teams, collecting management organisations (CMOs) and cultural institutions need to understand, provides step‑by‑step licensing pathways, and offers sample contract language that can be adapted for negotiation.

What Greek Businesses and Cultural Institutions Must Know Now

The short answer is that permission is generally required before copyrighted works governed by Greek law are incorporated into an AI training dataset. Neither Law 5179/2025 nor Law No. 5271/2026 creates a blanket exemption for machine‑learning ingestion, and the text‑and‑data‑mining exceptions transposed from the EU DSM Directive remain narrow in scope. Any entity, whether a start‑up, multinational or public museum, that feeds Greek‑protected works into a model without authorisation faces escalated administrative fines and potential civil liability.

The two statutes work in tandem. Law 5179/2025 raised the ceiling on administrative fines imposed by the Hellenic Copyright Organisation (OPI) and tightened enforcement mechanisms for unauthorised online uses, including the scraping and reproduction of protected content for AI purposes. Law No. 5271/2026, published in February 2026, added express provisions on cultural‑heritage and CMO‑managed works, mandating explicit consent for AI training uses and clarifying the powers of the competent administrative authorities.

Industry observers expect that enforcement activity will intensify throughout 2026 as OPI operationalises these new powers. Organisations should take three immediate steps: (1) pause any unverified dataset pipelines containing Greek‑sourced material, (2) conduct a rights‑provenance audit of existing training corpora, and (3) open licensing discussions with the relevant CMOs or individual rights holders.

Legal Framework: Baseline Copyright Law and Recent Amendments

Greece has maintained a comprehensive copyright regime since the adoption of Law 2121/1993 on copyright, neighbouring rights and cultural matters. This foundational statute grants authors an exclusive bundle of economic rights, reproduction, adaptation, distribution, public performance and communication to the public, alongside inalienable moral rights. It also establishes the framework for collective management and designates the Hellenic Copyright Organisation (OPI) as the supervisory authority over CMOs.

Between 2019 and 2021, Greece transposed the EU Directive on Copyright in the Digital Single Market (DSM Directive 2019/790). This transposition introduced limited text‑and‑data‑mining (TDM) exceptions, new neighbouring rights for press publishers, and updated rules on the liability of online content‑sharing platforms. Crucially, the TDM exception for commercial actors under Article 4 of the DSM Directive was transposed in a form that allows rights holders to reserve their rights through machine‑readable opt‑out mechanisms, a reservation many Greek publishers and CMOs have since activated.

The legislative pace then accelerated. Law 5179/2025 marked the first post‑DSM overhaul, raising administrative fine ceilings, broadening the scope of “unauthorised online use” to expressly encompass automated dataset collection for AI model training, and giving OPI enhanced investigative and injunctive powers. Months later, Law No. 5271/2026 refined these provisions with targeted measures for cultural‑heritage collections and works administered by CMOs, requiring explicit written permissions before such material can be used in AI training and imposing additional record‑keeping obligations on licensees.

Copyright Lawyers Greece: Permission Rules for AI Training Data

Under Greek law, training an AI model on copyrighted works engages several exclusive rights simultaneously. The ingestion step involves reproduction; feature extraction may constitute adaptation; and making the model or its outputs available online can trigger the right of communication to the public. Unless a statutory exception applies, each of these acts requires prior authorisation from the rights holder.

When Is Permission Required?

Permission is required whenever the work is still within its copyright term and no applicable exception covers the intended use. In Greece, the general copyright term is 70 years post mortem auctoris for works of authorship, and 50 years from first publication or fixation for neighbouring rights. If a work falls within these terms and the rights holder (or their CMO) has not granted a licence, training use is presumptively infringing.

The Text‑and‑Data‑Mining Exceptions

Greece’s transposition of the DSM Directive provides two TDM exceptions. The first, based on DSM Article 3, permits TDM by research organisations and cultural‑heritage institutions for scientific research purposes, provided they have lawful access to the source material. The second, based on DSM Article 4, permits TDM by any party, including commercial operators, but only where the rights holder has not opted out using appropriate technological measures. In practice, many Greek publishers, CMOs and cultural institutions have exercised this opt‑out right, rendering the commercial exception unavailable for a significant portion of Greek‑sourced content.

Law 5179/2025 narrowed the practical scope of the commercial TDM exception further by clarifying that automated scraping of online repositories, digital archives and streaming platforms for AI training purposes constitutes “unauthorised reproduction” unless the data user can demonstrate either (a) a valid licence or (b) confirmed absence of an opt‑out reservation. The burden of proof rests on the data user.

Permission Decision Flowchart

AI teams should follow this stepwise assessment before incorporating Greek‑sourced content into a training pipeline:

1. Check copyright status. Is the work still in its copyright term? If expired, proceed with caution, moral rights may still apply.
2. Identify the rights holder. Is the work administered by a CMO, held by a museum or archive, or controlled directly by the author or publisher?
3. Check for opt‑out reservations. Has the rights holder activated machine‑readable opt‑out tags (e.g., robots.txt directives, metadata flags)?
4. Determine whether a research exception applies. Are you a qualifying research organisation or cultural‑heritage institution conducting non‑commercial scientific research?
5. If no exception applies, obtain a licence. Contact the CMO for collective works or negotiate directly with the rights holder. Document every consent in writing.
6. Record provenance. Log the source, date of ingestion, licence reference and opt‑out check result for each dataset element.

Responding to Takedown or Enforcement Notices

If an AI developer receives an enforcement notice from OPI or a cease‑and‑desist letter from a rights holder, the recommended response sequence is: (1) preserve all relevant records and dataset logs, (2) suspend ingestion of the disputed material immediately, (3) engage qualified copyright counsel to assess the merits of the claim, and (4) respond within the statutory deadline set out in the notice, typically 15 business days for administrative proceedings.

CMOs and Collective Licensing: Negotiating Blanket Licences for AI Datasets

Collecting management organisations are the primary licensing gateway for large‑scale AI training projects that involve Greek‑protected works. CMOs aggregate the rights of individual authors and publishers, enabling data users to obtain a single blanket licence rather than negotiating thousands of individual permissions.

Key Greek CMOs

Greece has several CMOs, each covering different categories of works. The principal organisations supervised by OPI include:

• AEPI (Hellenic Society for the Protection of Intellectual Property). Manages musical works, both composition and lyrics, on behalf of Greek and international authors through reciprocal agreements.
• ATHINA (Authors’ Society). Represents literary and dramatic works, including prose, poetry and theatrical scripts.
• OSDEETE. Manages rights for audiovisual works and screenwriters.
• OSDEL. Covers reprographic and digital reproduction rights for publishers and authors of written works.
• GRAMMO / ERATO. Administer neighbouring rights for phonogram producers and performing artists respectively.

Law No. 5271/2026 requires CMOs to establish transparent licensing terms for AI training uses and to publish tariff schedules on their websites. Early indications suggest that most Greek CMOs are developing dedicated AI training licence categories, though standardised pricing had not been finalised across all organisations as of the date of this guide.

How to Obtain a Blanket CMO Licence

The process typically involves four stages: (1) identify which CMO(s) administer the rights relevant to your dataset, (2) submit a formal licence application describing the scope, volume and purpose of the intended training use, (3) negotiate fees based on the tariff schedule and dataset size, and (4) execute a written licence agreement that includes data‑retention limits, usage reporting obligations and audit rights. Based on current practice, industry observers expect the end‑to‑end process to take between eight and sixteen weeks from initial application to executed agreement.

Limitations on CMO Licensing

CMOs cannot waive or licence moral rights on behalf of their members. Greek moral rights, including the right of attribution and the right of integrity, are inalienable under Law 2121/1993. This means that even with a valid blanket economic licence, the data user must ensure proper attribution in model documentation and must not distort the work in a manner that harms the author’s reputation. Additionally, certain categories of visual art and photographic works may not be collectively managed; direct negotiation with the artist or their estate is often necessary.

Cultural‑Heritage Materials and Museums: Special Rules for AI Training

Law No. 5271/2026 introduced heightened protections for cultural‑heritage materials held by Greek museums, archives and public collections. These provisions respond to concerns that mass digitisation and AI training could exploit irreplaceable national‑heritage works without adequate safeguards.

What the New Rules Require

Under the February 2026 measures, any use of cultural‑heritage materials for AI model training requires explicit written consent from the custodial institution, even where the underlying copyright has expired. This consent must specify the permitted uses, the duration of the licence, attribution requirements, and the data user’s obligation to delete training copies upon licence expiry. The custodial institution retains the right to withdraw consent if the AI outputs are deemed to misrepresent or decontextualise the heritage work.

Orphan Works and Unresolved Authorship

Many cultural‑heritage collections include orphan works, pieces for which the rights holder cannot be identified or located after a diligent search. Greece’s orphan‑works framework, derived from EU Directive 2012/28/EU, permits certain uses by publicly accessible cultural institutions but does not extend to commercial AI training. For orphan works intended for AI datasets, the recommended approach is to conduct and document a diligent search, register the work with EUIPO’s orphan‑works database, and apply to OPI for an administrative licence where available.

Moral Rights and Sensitive Materials

Cultural‑heritage works frequently carry heightened moral‑rights sensitivities. Sacred objects, works of indigenous or minority‑community origin, and politically sensitive archival material may require additional ethical review beyond legal copyright clearance. Industry observers expect the Greek Ministry of Culture to issue supplementary guidance on ethical AI use of heritage collections in the second half of 2026. Museums should build an internal review process that considers both legal permissions and ethical stewardship before releasing any collection material into AI pipelines.

Enforcement, Administrative Fines and Responding to Notices

The enforcement regime for copyright infringement in Greece now combines civil remedies, criminal sanctions and a significantly strengthened administrative fines framework. Law 5179/2025 empowered OPI to impose administrative fines on entities found to have engaged in unauthorised reproduction, communication or dataset collection of protected works. Law No. 5271/2026 extended these powers to cover cultural‑heritage materials specifically.

Administrative Fines Under Law 5179/2025 and Law No. 5271/2026

How to Respond to an Enforcement Notice

Entities that receive an administrative notice or cease‑and‑desist letter should follow this response protocol:

1. Do not ignore the notice. Administrative fines can be imposed in default if no response is filed within the stated deadline.
2. Preserve all evidence. Secure dataset logs, ingestion records, licence agreements and opt‑out check documentation.
3. Engage specialist copyright counsel. Copyright lawyers Greece‑based will assess whether a valid defence (e.g., applicable TDM exception, valid licence) exists.
4. File a written response. Address each allegation factually; attach supporting evidence.
5. Consider settlement. Where liability is clear, early engagement with the rights holder or CMO to agree remedial licensing and reduced penalties is often the most cost‑effective path.
6. Appeal if necessary. Administrative fines imposed by OPI may be challenged before the competent Administrative Court of Appeal within the statutory appeal period.

Litigation risk remains significant. Beyond administrative fines, rights holders retain the right to pursue civil damages and injunctive relief through the Greek courts. In cases involving large‑scale commercial infringement, criminal prosecution remains a possibility, carrying potential imprisonment of up to ten years for the most serious offences under Law 2121/1993.

Practical Compliance Checklist and Sample Licence Language

The following checklist is designed for AI teams, museum compliance officers and in‑house counsel seeking to achieve and maintain copyright compliance for AI developers operating in or with Greek‑sourced content.

12‑Point Compliance Checklist

1. Dataset audit. Catalogue every Greek‑sourced work in your training corpus by title, author, source and date of ingestion.
2. Rights mapping. For each work, determine whether it is in copyright, who the rights holder is, and whether a CMO administers the relevant rights.
3. Opt‑out check. Verify whether the rights holder has activated machine‑readable opt‑out tags; document results with timestamps.
4. Exception assessment. Evaluate whether the research TDM exception (DSM Art. 3) applies to your organisation and use case.
5. CMO contact. Initiate blanket licence discussions with the relevant CMO(s) for collectively managed works.
6. Direct licences. For works not covered by CMOs, negotiate individual licences with rights holders or their representatives.
7. Cultural‑heritage clearance. For museum or archival material, obtain written consent from the custodial institution per Law No. 5271/2026.
8. Moral‑rights compliance. Ensure attribution is preserved in model documentation and that AI outputs do not distort the integrity of the source works.
9. Provenance metadata. Tag every dataset element with provenance, licence reference, opt‑out status and ingestion date.
10. Record retention. Maintain complete provenance and consent records for a minimum of ten years, the limitation period for civil copyright claims in Greece.
11. Redaction policy. Establish a documented process for removing works from the dataset upon licence expiry, rights‑holder withdrawal or enforcement notice.
12. Periodic re‑audit. Schedule rights audits at least annually to capture new opt‑out reservations, licence expirations and legislative changes.

Sample Licence Clauses (For Negotiation Only)

Disclaimer: The following clauses are provided as starting points for negotiation and do not constitute legal advice. Organisations should engage qualified copyright lawyers in Greece to adapt these templates to their specific circumstances.

Clause 1, Blanket CMO Licence for AI Training

“The CMO hereby grants the Licensee a non‑exclusive, non‑transferable licence to reproduce and adapt Works within the CMO’s Repertoire solely for the purpose of training the Licensee’s specified AI Model(s). The Licensee shall (a) maintain a complete log of all Works ingested, (b) submit quarterly usage reports to the CMO, and (c) delete all training copies within 30 days of licence expiry or termination.”

Clause 2, Museum/Cultural‑Heritage Institution Dataset Licence

“The Institution grants the Data User a limited, revocable licence to use digitised reproductions of Collection Items for AI model training, subject to the following conditions: (i) training use only, no reproduction in model outputs; (ii) full attribution in model documentation; (iii) immediate takedown of any Collection Item upon written notice from the Institution; (iv) the Data User shall indemnify the Institution against third‑party claims arising from the AI use of Collection Items.”

Clause 3, Developer Indemnity and Audit

“The Developer warrants that all works incorporated into the Training Dataset have been licensed or fall within a valid statutory exception. The Developer shall indemnify the Licensor against all claims, fines and costs arising from unauthorised use. The Licensor may, upon 10 business days’ written notice, audit the Developer’s dataset logs and provenance records to verify compliance.”

When to Engage a Copyright Lawyer

Not every AI project requires external counsel, but several triggers should prompt immediate engagement with copyright lawyers Greece‑based or with cross‑border IP expertise. These triggers include: assembling a large training dataset from Greek sources (more than 1,000 works), any use of cultural‑heritage materials subject to Law No. 5271/2026, commercial launch of an AI product trained on Greek‑protected content, receipt of an enforcement notice from OPI or a rights‑holder cease‑and‑desist, and any cross‑border scenario where Greek copyright interacts with foreign IP regimes.

A retained copyright lawyer will typically conduct a dataset rights audit, negotiate CMO and direct licences, draft or review training‑data licence agreements, advise on enforcement response strategy, and represent the organisation in administrative or civil proceedings. Organisations seeking qualified Greek IP counsel can browse the Global Law Experts Greece lawyer directory or consult the international intellectual property guide for cross‑border matters.

Sources

1. Hellenic Copyright Organisation (OPI), Official Guidance and CMO Supervision
2. Government Gazette (Εφημερίδα της Κυβερνήσεως), Laws 5179/2025 & 5271/2026
3. EUIPO, Copyright and AI Resources
4. ICLG, Greece Copyright Chapter
5. Wolters Kluwer / Kluwer IP Blog, EU AI and Copyright Analysis
6. Legal 500, Greece Intellectual Property Rankings
7. Greek Ministry of Culture, Museums and Cultural Heritage Guidance