Denmark's Minister of Culture, Zenia Stampe, on October 8, 2026 introduced bill L 49 in the Folketing, a proposal that would make it unlawful to make a realistic, digitally generated imitation of any person's appearance, voice or movements available to the public without that person's consent. The 36-page bill, which would amend the Danish Copyright Act from January 1, 2027, places the burden of proving consent on whoever publishes the imitation, protects individuals for 50 years after death and gives courts a new power to order unlawful content removed from the platform where it appears. It is the second attempt: according to Copenhagen lawyer Jakob Plesner Mathiasen, an earlier version lapsed when Denmark went to a general election.
In Short
Denmark wants a law that says nobody may post a realistic fake video, image or voice clip of you unless you agreed to it. That matters for anyone who uses AI to make ads, music or social media content featuring real people, because the person who publishes the fake would have to prove they had permission. Jokes, satire and parody stay allowed, but not when a fake is realistic enough to mislead people in a way that could cause serious harm, and the rules would start on January 1, 2027 if parliament passes them.
A second attempt at the same idea
The bill carries the number L 49 for the 2026-27 parliamentary session and was "Fremsat den 8. oktober 2026 af kulturministeren (Zenia Stampe)" - tabled on October 8, 2026 by the Minister of Culture - according to the document's cover page. Its full title is a proposal for an act amending the Copyright Act, with the subtitle "protection of personal characteristics etc. and amended procedure for review by the Copyright License Tribunal."
The policy itself is older than the bill. According to the explanatory notes, L 49 implements a political agreement of June 26, 2025, concluded between "the then SVM government" - the coalition of the Social Democrats, Venstre and the Moderates - and six other parties: the Denmark Democrats, the Socialist People's Party, the Conservative People's Party, the Red-Green Alliance, the Social Liberals and the Alternative. The phrase "the then" is itself a marker of how Danish politics has moved since. A draft was sent for public consultation between July 7 and August 21, 2025, a period of 45 days, and a separate draft on the tribunal procedure went out between December 10, 2025 and January 9, 2026, a period of 30 days.
Mathiasen, a partner at Gorrissen Federspiel who heads the firm's Media and Entertainment division, explained the gap in a LinkedIn post published the day the bill was tabled. "Today, the Danish Minister of Culture reintroduced a bill aimed at protecting individuals against digital deepfakes," Mathiasen wrote. "The original bill lapsed following the Danish general election. The new bill has now been introduced in a revised form."
That history has caused confusion. In the comments beneath Mathiasen's post, one trademark specialist wrote that they had thought the measure "had been passed as legislation on March 31," and recalled that people last year had been making "LOTS of NOISE of APPRECIATION" on the assumption it was already law. It was not, and it still is not. Another commenter asked whether an English translation existed. "As far as I know, the bill is currently only available in Danish," Mathiasen replied. The quotations from the bill in this article are translated from the Danish original.
Denmark has moved early on AI rules before. The country became the first EU member state to adopt national legislation implementing the AI Act, passing it on May 8, 2025, ahead of the August 2, 2025 deadline.
What section 73 a would prohibit
The core of the bill is a new section 73 a in the Copyright Act. In the bill's words: "Realistic digitally generated imitations of a natural person's personal, physical characteristics may not be made available to the public without the consent of the imitated person."
Four elements in that sentence carry most of the legal weight, and the explanatory notes define each.
Realistic. According to the notes, an imitation is realistic when it is "suitable to give rise to an obvious risk of confusion" with an actual depiction of the person. The ministry offers unusually concrete examples of what falls outside. A person rendered as a fantasy creature would generally not be covered. Neither would a likeness placed on an alien, or an otherwise lifelike figure with "neon green skin and pink eyes." But subtle, non-realistic features such as "discreet fangs or elf ears" may not be enough to escape the rule. A well-executed, digitally generated portrait drawing or painting is also excluded, on the reasoning that it could not be mistaken for reality because of its style. Imitations placed in a context that is explicitly a recognisable fantasy world fall outside as well.
Digitally generated. The notes say this term "must be understood broadly." It covers imitations made with "sophisticated technological aids such as artificial intelligence" and those made with "less sophisticated technological editing tools." Mathiasen made the same point in a summary of the bill: "The provision covers both AI generated deepfakes and imitations created using more traditional digital editing tools." The bill states it does not intend to alter "ordinary industry norms" for technical measures used in recording or editing film, sound and images.
Personal, physical characteristics. These are defined as features unique to an individual, such as appearance, voice and movements, which together make a person recognisable. Individual elements can be protected on their own if an imitation of that element alone creates an obvious risk of confusion. The notes give characteristic voices as an example. The provision protects only natural persons: company names and trademarks are expressly outside its scope.
Made available to the public. The bill borrows this term from section 2 of the Copyright Act, which follows EU law. Creating a deepfake is not, by itself, unlawful. According to the notes, realistic imitations could still be shared "within the private sphere, for example at a private party." Mathiasen summarised it this way: "Only making the imitation available to the public is covered. Private uses are generally excluded."
There is also a narrow exception for trivial uses. Where a realistic imitation is "of very insignificant character in relation to the larger context" in which it appears - a film production is the example - the notes say it falls outside the rule. That exception, they add, "must be interpreted restrictively."
No intent required, and no criminal penalty
Two features separate section 73 a from Denmark's existing criminal provisions on identity misuse. According to the explanatory notes, the new rule does not require that the imitation be "improper," which section 264 e of the Criminal Code does. And it does not require intent, meaning "the provision therefore also applies to accidental and negligent publication without consent."
At the same time, breaching section 73 a would not be a crime. The bill amends section 76 of the Copyright Act specifically so that the new provision falls outside its fines. Redress would run through the civil courts, where, according to the notes, liability for losses and compensation for non-pecuniary harm could be imposed. Mathiasen addressed this directly when a commenter argued the law would only create "more work for the Police." The reply: "One important aspect of the proposal is that it provides individuals with a clearer legal basis for taking civil action against unauthorised digital imitations, including seeking their removal. So enforcement is not necessarily a matter for the police."
Consent that can be withdrawn, and a burden of proof that sits with the publisher
The bill's consent standard is drawn closely from data protection law. According to the notes, consent exists when there is "a voluntary, specific, informed and unambiguous declaration of will" by which the imitated person agrees, by statement or clear affirmative action, that a digitally generated imitation may be made public.
Consent can be withdrawn at any time, with no formal requirements for doing so. Withdrawal does not make earlier publication unlawful; it affects only future publication. Where consent forms part of a contract, withdrawal follows ordinary contract law and does not by itself void or end the contract.
Then comes the sentence most relevant to anyone producing commercial content: "It is the person who makes the realistic digitally generated imitation available to the public who must prove that consent from the imitated person exists." Mathiasen listed it as the ninth of nine key points: "The person making the imitation available to the public bears the burden of proving that consent was obtained."
Consent, and the right to subsequent use of an imitation, can be transferred by agreement, according to the notes. But not through collective deals. The bill states that personal characteristics are so closely tied to the individual that the area cannot be covered by an agreement with extended collective licensing effect. In Mathiasen's words, collective management organisations "cannot license these personal characteristics on behalf of individuals through extended collective licensing without individual consent."
Fifty years after death, and anyone in the world
Protection under section 73 a would last until 50 years after the end of the year in which the imitated person died. The notes justify the figure on two grounds: an imitation released soon after a death could be offensive to the bereaved, while protection without end would intrude too far on freedom of expression. Fifty years, the ministry concluded, is sufficient to prevent a digital imitation from causing "significant damage to the perception of history in the society in which it is made available."
Mathiasen singled this out. "This is particularly interesting because the protection of deceased individuals against (digital) imitations has been a difficult legal question internationally," Mathiasen wrote. "In many jurisdictions, the scope and duration of protection remain unclear and often depend on case law."
The territorial scope is broad. A new section 86 a would apply section 73 a to "all natural persons, including foreign nationals." According to the notes, the scope should apply regardless of whether a person is a citizen of Denmark or of an EU or EEA country. Ordinary rules on choice of law and jurisdiction apply.
Satire, parody and the misinformation exception
Section 73 a(2) carves out imitations that are "an expression of caricature, satire, parody, criticism of power, social criticism or similar." The exception has its own exception: it does not apply where the imitation "constitutes misinformation which can concretely cause serious danger to the rights or significant interests of others."
The notes define caricature and its relatives as imitations that are "a clear expression of humour, entertainment or ridicule," using devices such as irony, sarcasm, obvious exaggeration or understatement. Misinformation means "untrue, misleading or erroneous information" that could cause potential danger or harm to others' life, reputation, health or property.
The ministry's examples are pointed. One is a realistic deepfake of Denmark's National Police Commissioner encouraging vigilantism. Another is a parody of a prominent politician, senior civil servant or public figure accusing a named person or group of a crime, where it is not clear the content is parody. A third is directly relevant to advertising: according to the notes, it could constitute misinformation endangering others' property if an image is published without consent showing a realistic imitation of a well-known person "calling for specific significant financial dispositions to be made."
That description fits the investment scam advertisements that have plagued social platforms for years. On September 16, 2026, the Frankfurt Regional Court ruled that Meta could not rely on the hosting provider defence under the Digital Services Act in a case where fraudsters had used photos of Finanzfluss co-founder Thomas Kehl in paid Facebook and Instagram ads for fraudulent investments, with fines of up to 250,000 euros per future breach. Earlier in the year, Meta filed lawsuits against scam advertisers in Brazil, China and Vietnam over celebrity impersonation, including a Brazilian network that used deepfakes of a physician.
On the reputational side, the notes set a high bar. Caricatures must create a risk of "particularly significant harm," such as the imitated person losing their job or income, or being linked to illegal or widely condemned acts.
Two further details would affect how short-form video is judged. For content with both picture and sound, the notes say it must be clear from both the image and the audio that the content is satire, because videos on social media "are often viewed without sound," or the audio is heard without the image. And whether something counts as misinformation "will not depend on" whether the person who published it was aware of that at the time.
There is a tension the ministry acknowledges. Its own general notes concede that, under the proposed rule, protection "will in principle cover" parodic and satirical imitations made without consent if they create an obvious risk of confusion. The bill then argues that the European Court of Human Rights' case law - which holds that politicians and heads of state must tolerate more criticism - "does not specifically take into account the risk of confusion with reality" that realistic digital imitations create. The ministry concluded that the ban can be implemented within Article 10 of the European Convention on Human Rights, while noting that the Strasbourg court does not appear to have ruled on realistic digital imitations directly.
News reporting
A fourth subsection allows previously published imitations to be reproduced without consent as part of news reporting, in line with sections 23(2) and 25 of the Copyright Act. According to the notes, that access applies only where reproduction is "a natural and necessary part of news coverage" - not where an imitation is used to illustrate information, as entertainment, or for other exploitation that is not necessary for news relevance.
A separate right for performers and artists
Alongside the general protection, the bill creates a second, narrower right in a new section 65 a. Realistic digitally generated imitations of a performer's rendition of a literary or artistic work, or of an artist's artistic performance, may not be made available to the public without the performer's or artist's consent.
This right is closer to copyright. Where section 73 a protects the person, section 65 a protects the performance. The notes give examples: a digitally generated imitation of an author's book reading would be covered, a use case that sits squarely within the fast-growing market for AI narration. So would an actor's performance used to generate a realistic imitation of the actor's face and voice in a new film scene with different content, published on an online content-sharing service. So would a song recording used to generate a realistic imitation of a singer's performance released via a music streaming service.
The ministry points to the music market specifically. According to the notes, original artists increasingly find that social media and music streaming services contain "more and more tracks that contain digital 1:1 imitations of musicians' vocals," which compete directly with the original music and can be offensive or discrediting. The bill states it is not aimed at legitimate cover recordings, and look-alike or sound-alike bands performing live fall outside, because they are not digitally generated.
Duration under section 65 a mirrors existing performers' rights: 50 years from the end of the year of the performance, extended to 50 years from first publication for recordings other than sound recordings, and 70 years for sound recordings, where publication happens within the first 50 years. The right applies to performers and artists who are nationals of, or resident in, a country in the European Economic Area - a narrower scope than the worldwide coverage of section 73 a.
Collective licensing is allowed here
Unlike section 73 a, the performers' right can be managed collectively. The bill applies the Copyright Act's extended collective licensing provisions to section 65 a, on the reasoning that individual clearance would become "disproportionately difficult" for mass use, as it already is for existing performers' rights. A collective management organisation could therefore agree payment terms with users on behalf of performers for use of their performances in realistic digital imitations. Disputes over those terms would fall under the Copyright License Tribunal.
Infringements of section 65 a would entitle the performer to reasonable remuneration and damages under section 83 of the Copyright Act, as well as compensation for non-economic harm. As with section 73 a, criminal penalties would not apply.
Artists who do not perform "works"
The bill also widens who counts as protected. Danish law has, since a 1961 report, protected performers only when they render a literary or artistic work. Acrobats, magicians, impersonators and variety or circus artists have generally fallen outside. L 49 would insert "or artists" throughout the Copyright Act, extending protection to "artistic performances" that are not performances of a work.
The ministry draws a line at sport. According to the notes, performances that are merely expressions of "technical skill, endurance or tactics," and sports performances with a competitive element aimed at an audience, would not be covered. The notes also address AI-generated material directly: there is no consensus in the literature, and no national case law, on whether computer-generated output can be a "work," so it is unclear whether a performance of an AI-generated song or play would be protected under the current section 65. The new artist category is intended to close that gap.
Removal from platforms
For enforcement, the bill extends section 84 of the Copyright Act, which already lets courts order infringing copies to be recalled, removed from commerce, destroyed or handed over, to cover sections 65 a and 73 a. It also adds a new remedy: courts could order that infringing material "be removed from the medium on which it has been made available."
The notes define "medium" as digital and analogue devices and platforms on which data can be stored or displayed, and list "social media platforms, online content-sharing services and file and document sharing platforms." The same paragraph then states that the term does not include intermediary services performing mere conduit or caching, or "a medium which constitutes a hosting service" as defined in Article 3(g) of the Digital Services Act. Read literally, the two sentences sit uneasily together, since social media platforms are a category of hosting service under the EU regulation. Elsewhere, the general notes state that the new rules must be read alongside the Digital Services Act's provisions on the obligation of hosting services to act against illegal content. Clarification may come during committee scrutiny.
Platforms are well aware of the issue. Google Denmark, Facebook, Instagram, LinkedIn, TikTok, Spotify and Netflix were among more than 100 bodies consulted on the 2025 draft, according to the bill's consultation list, and Meta was on the list for the second consultation. Advertising and marketing bodies were consulted too, including Danske Annoncører og Markedsførere, the Danish association of advertising agencies, and, in the second round, Dansk Annoncørforening.
Some platforms have built their own systems in the meantime. In August 2026, YouTube added four dispute categories for AI likeness claims, including explicit consent and "parody, satire, or public interest" - categories that closely resemble the exceptions in the Danish bill. Its likeness detection tool, initially limited to Partner Program members, was opened to all eligible creators aged 18 and over in May 2026 after a pilot for politicians, government officials and journalists in March. A commenter on Mathiasen's post, Jakub Wyczik, predicted that "platforms will continue to rely on the safe harbor hosting provisions," and that effective action against anonymous infringers will be needed.
How the bill sits next to the AI Act
The bill is careful to separate itself from EU rules. According to its notes, L 49 "does not contain provisions that implement EU law obligations." The performers' right "is a purely national provision," even though it uses the EU-harmonised concept of making available to the public. Because it may regulate digital services, the bill was notified to the European Commission under Directive (EU) 2015/1535, the technical standards notification procedure.
The notes describe the overlap with the AI Act at some length. Article 50(2) requires providers of AI systems generating synthetic audio, image, video or text to mark output in a machine-readable format, the job that provenance standards such as C2PA were designed to do. Article 50(4) requires deployers of AI systems that generate or manipulate deepfake content to disclose that it has been artificially generated or manipulated. The bill quotes the AI Act's definition of a deepfake as AI-generated or manipulated image, audio or video content that "appreciably resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful." It also notes that the digital omnibus on AI has introduced "two new prohibitions against sexualising deepfakes."
The relationship is explicit: section 73 a does not change obligations under the AI Act, and "any consent from an imitated person in accordance with this act does not override those obligations." A labelled deepfake made with consent would still need to be labelled; a labelled deepfake made without consent would still be unlawful in Denmark.
For marketers, the two regimes now stack. Article 50 obligations have applied since August 2, 2026, and the Commission's Code of Practice and guidelines, published on July 20, 2026, set fines of up to 15 million euros or 3 percent of worldwide annual turnover. An IAB Austria guide from September 2026 concluded that agencies, not clients, usually carry the label duty as the deployer of the AI tool. Google, for its part, added a "How this ad was made" panel on July 9, 2026, indicating whether generative AI was used to create or edit ads.
Disclosure answers the question of whether content is synthetic. The Danish bill answers a different one: whose likeness it uses, and whether they agreed.
What Danish law already covers
The ministry is clear that the new provisions supplement rather than replace existing rules. According to the notes, the Danish Marketing Practices Act already makes it contrary to good marketing practice to use a person's image in marketing without prior permission. A separate executive order on advertising in radio, television and on-demand services prohibits ads that depict or refer to people without prior permission. Two unwritten principles of Danish law, set out in Supreme Court rulings of January 25, 1965 (the Buster Larsen case) and June 10, 2010, already restrict the commercial use of a person's image and allow compensation for unlawful publication.
What changes is reach. According to the notes, section 73 a differs from the marketing rules "by also applying to non-commercial" publication of digitally generated imitations. Mathiasen made a similar distinction in reply to Wyczik's comment that the bill amounts to "the Danish version of the right of publicity." The answer: "There are certainly similarities with the right of publicity, although the proposed Danish protection is not limited to commercial exploitation."
Germany reached a comparable result through the courts. In August 2025, the Berlin Regional Court ruled that a YouTuber's use of an AI-cloned voice of a professional voice actor infringed the actor's personality rights, ordering 4,000 euros in damages, and held that a licence from the AI tool provider did not replace the actor's own consent. Italy has taken a criminal route: a decree published in September 2026 created corporate liability for unlawful dissemination of AI-generated content, with sanctions that can include a ban on advertising for up to two years. And on October 5, 2026, Italian Prime Minister Giorgia Meloni applied to register a four-second recording of the prime minister's own voice as an EU sound trademark, with the prime minister's office citing protection against deepfakes - an indication of how far individuals are reaching for tools the law does not yet provide.
The tribunal changes
The bill's second subject receives less attention but would change how licensing disputes are handled. The Copyright License Tribunal, set up in 1961 and chaired by a Supreme Court judge with two other members, rules on disputes over remuneration and other terms in collective and compulsory licences. Under current law, parties can go to the tribunal and the courts in parallel.
A new section 47(3) would end that. As long as the tribunal is handling a question on the reasonableness of remuneration or licence terms, the parties could not bring the same question before the courts. If a case has already been filed in court and one party asks for tribunal treatment, the court would have to dismiss the case and send it to the tribunal. Requests would have to be made at the preparatory hearing at the latest, or within 4 weeks of the court's notice that no such hearing will be held. Tribunal decisions could still be challenged in court afterwards. The bill also gives the minister power to set rules for digital communication with the tribunal, and corrects a 2024 drafting error under which section 47(2) omitted the tribunal's competence over section 67 a.
Discrepancies in the text
The bill as tabled contains several internal inconsistencies. The transitional provision in section 2(2) of the bill text excludes "§ 1, nr. 7 og 9" - the points creating sections 65 a and 73 a - from agreements made before entry into force. The explanatory note to the same provision refers instead to "§ 1, nr. 9 og 11," although point 11 concerns an unrelated amendment to section 77; the note's following sentence confirms that the intended targets are sections 65 a and 73 a. The notes to section 73 a also refer to the performers' right as "lovforslagets § 1, nr. 10," when it is point 7. And the list of Copyright Act provisions applied to section 65 a in subsection 5 differs between the bill text and the explanatory notes. The general notes describe the protection as covering "very realistic" imitations in some headings and "realistic" imitations in the operative text. None of these alters the substance of the proposal, but they are the kind of points that tend to be corrected in committee.
Why this matters for advertising
The bill's own impact assessment says little about business costs. According to section 5 of the notes, the economic consequences for business are expected to be "limited," arising mainly from "any expenses for payment for exploitation of the new right" under section 65 a. No administrative consequences for business are listed.
That assessment covers the performers' right only. Section 73 a has wider practical reach. Any advertiser, agency, creator or publisher that makes a realistic AI-generated or digitally edited image, video or voice of a real person available to the public in Denmark would need that person's consent - and would need to be able to prove it. The rule would apply to non-commercial content as well as advertising, to private individuals as well as celebrities, and to people who died up to 50 years ago. It would cover synthetic voice-overs that imitate a recognisable voice, generated spokesperson videos, AI-edited influencer content and localised versions of campaigns in which a person's face or speech has been altered.
The withdrawal rule adds an operational question. A person who consented could withdraw consent for future publication at any time, without formalities, although contracts would remain governed by ordinary contract law. Campaign assets featuring a realistic imitation could therefore need to be taken down if consent is withdrawn and the contract does not settle the point.
The timing gives some room. Under the transitional rule, sections 65 a and 73 a would not apply to agreements concluded before January 1, 2027. Agreements concluded or extended after that date would be assessed under the new rules.
For platforms, the bill does not create new filtering duties. It gives individuals a statutory basis to demand removal, backed by a court power to order material taken down. "In short, the proposal provides a clearer legal basis for individuals to act against unauthorised deepfakes, including seeking the removal of unlawful content," according to Mathiasen.
The Folketing must still debate and pass the bill before the January 1, 2027 date can hold. The law would not apply to the Faroe Islands, which took over copyright legislation in 2010, or to Greenland, although it could be brought into force there by royal decree.
Timeline
- 1961 - Denmark establishes the tribunal now known as the Copyright License Tribunal
- January 25, 1965 - Danish Supreme Court rules in the Buster Larsen case that a person's image may not be used commercially without permission
- June 10, 2010 - Danish Supreme Court ruling on compensation for unlawful publication of a person's image
- May 8, 2025 - Denmark becomes the first EU member state to pass national AI Act implementing legislation
- June 26, 2025 - Danish government and six other parties agree to create protection against sharing digital imitations
- July 7 to August 21, 2025 - First draft of the deepfake bill sent for a 45-day consultation
- August 20, 2025 - Berlin Regional Court rules AI voice cloning of a voice actor violates personality rights
- December 10, 2025 to January 9, 2026 - Draft changes to the Copyright License Tribunal procedure sent for a 30-day consultation
- February 26, 2026 - Meta files lawsuits against celeb-bait scam advertisers in Brazil, China and Vietnam
- March 10, 2026 - YouTube extends likeness detection to politicians, officials and journalists
- May 2026 - YouTube opens likeness detection to all eligible creators aged 18 and over
- July 9, 2026 - Google adds "How this ad was made" AI disclosure panel
- July 20, 2026 - European Commission publishes AI Act Article 50 Code of Practice and guidelines
- August 2, 2026 - AI Act Article 50 transparency obligations apply
- August 19, 2026 - YouTube adds four dispute categories for AI likeness claims
- September 15, 2026 - Italy publishes decree creating corporate liability for unlawful AI content dissemination
- September 16, 2026 - Frankfurt Regional Court rejects Meta's hosting defence over fake Finanzfluss ads; IAB Austria publishes AI ad labelling guide
- October 5, 2026 - Giorgia Meloni applies to register a four-second voice recording as an EU sound trademark
- October 8, 2026 - Culture Minister Zenia Stampe tables bill L 49 in the Folketing; Jakob Plesner Mathiasen publishes an analysis on LinkedIn
- January 1, 2027 - Proposed entry into force of L 49
Related PPC Land coverage
- Denmark sets precedent with early AI Act implementation legislation - Denmark's May 2025 law naming three authorities to enforce the AI Act.
- German court rules AI voice cloning violates personality rights - The Berlin ruling that an AI tool licence does not replace a voice actor's consent.
- German court ends Meta's hosting shield, 250,000 euros per fake ad - The Frankfurt decision on scam ads using a financial educator's photos.
- Meta sues scam advertisers in Brazil, China and Vietnam over celeb-bait and cloaking - Lawsuits against networks using celebrity and physician deepfakes.
- EU AI content rules force publishers to label or risk 3% of turnover - The Article 50 Code of Practice, marking thresholds and fines.
- Agencies, not clients, usually carry AI label duty, IAB Austria guide says - Who counts as the deployer when AI is used to make ads.
- Google adds How this ad was made AI labels 24 days before EU deadline - Google's ad-level AI disclosure across Search, YouTube and Discover.
- YouTube adds 4 dispute categories for AI likeness claims that block videos - Consent and parody disputes for likeness complaints in YouTube Studio.
- YouTube's AI face shield now covers every creator 18 and over - The expansion of YouTube's likeness detection beyond the Partner Program.
- YouTube expands likeness detection to politicians and journalists - will it hold? - The March 2026 pilot for public figures.
- Italian firms face ad bans over AI-generated content offences - Italy's decree on AI offences and corporate sanctions.
- Meloni files 4-second recording of herself as an EU sound trademark - A trademark filing pitched as protection against deepfakes.
- Brazil bans paid ads for deepfakes in final 72 hours before elections - Brazil's data protection authority report on deepfake fraud and election rules.
Summary
Who: Denmark's Minister of Culture, Zenia Stampe, who tabled bill L 49; the Danish Folketing, which must pass it; and Jakob Plesner Mathiasen, a partner at Copenhagen law firm Gorrissen Federspiel, who analysed the bill on LinkedIn. The rules would affect anyone who publishes realistic digital imitations of real people, including advertisers, agencies, creators, platforms, performers and artists.
What: A bill amending the Danish Copyright Act to prohibit making realistic, digitally generated imitations of any natural person's appearance, voice or movements available to the public without consent (section 73 a), with exceptions for satire and parody unless the content is harmful misinformation. It adds a separate imitation right for performers and artists (section 65 a) that can be licensed collectively, puts the burden of proving consent on the publisher, protects individuals for 50 years after death, and lets courts order removal from the medium where content was published. It also requires licensing disputes to go to the Copyright License Tribunal first when one party requests it.
When: The bill was tabled on October 8, 2026, implementing a political agreement of June 26, 2025. An earlier version lapsed after the Danish general election, according to Mathiasen. The proposed date of entry into force is January 1, 2027, with agreements concluded before then unaffected.
Where: Denmark, excluding the Faroe Islands and Greenland. The personal-characteristics protection would cover imitations of all natural persons, including foreign nationals, while the performers' right covers nationals and residents of the European Economic Area.
Why: According to the bill, easy-to-use tools now make it "soon impossible" to tell real from manipulated material, creating risks for democratic debate and for performers whose voices and appearances can be copied. Existing Danish rules covered commercial use of a person's image, identity misuse and defamation, but did not give every individual a clear civil basis to have realistic digital imitations removed. For marketers, the bill adds a consent requirement on top of the AI Act's labelling duties: disclosing that an ad is AI-made would not be enough if it uses a real person's likeness without permission.
Discussion