{"id":3910,"date":"2026-07-27T09:33:39","date_gmt":"2026-07-27T09:33:39","guid":{"rendered":"https:\/\/sawaryn.com\/?p=3910"},"modified":"2026-10-23T10:53:43","modified_gmt":"2026-10-23T10:53:43","slug":"ai-voice-theft-poland-marketing-risks","status":"publish","type":"post","link":"https:\/\/sawaryn.com\/en\/publikacje\/ai-voice-theft-poland-marketing-risks\/","title":{"rendered":"AI Voice Theft: Poland\u2019s First Lawsuit and What Marketers Must Change"},"content":{"rendered":"

In Poland, there is no provision that says in so many words, \u201cYour voice is protected.\u201d Yet in November 2025, the first Polish lawsuit over AI voice theft<\/strong> was filed with the Warsaw Regional Court. The defendant may lose on three fronts at once: civil, administrative, and criminal.<\/p>\n

If your company uses AI voice generators in advertisements, commercials, onboarding, or chatbots, this article concerns you. Below is an analysis of the Jaros\u0142aw \u0141ukomski case, three areas of legal risk under the Civil Code, GDPR, and Criminal Code, and concrete steps to take before the problem affects your organization.<\/p>\n

The \u0141ukomski case is not an abstract case study. It signals that Polish courts will decide deepfake-audio cases, and the case law now taking shape will affect every company using synthetic voices.<\/p>\n

Basic terms you need to know<\/h2>\n

Before we begin the analysis, here are a few terms used in the article:<\/p>\n\n\n\n\n\n\n\n\n\n
Term<\/th>\nMeaning<\/th>\n<\/tr>\n<\/thead>\n
Audio deepfake<\/strong><\/td>\nAn audio recording generated or manipulated by AI to imitate a specific person\u2019s voice<\/td>\n<\/tr>\n
Voice likeness<\/strong><\/td>\nA concept in Polish legal scholarship: a person\u2019s voice treated as a protected personal right when listeners can recognize it<\/td>\n<\/tr>\n
Biometric data<\/strong><\/td>\nPersonal data resulting from specific technical processing and relating to a person\u2019s physical or behavioral characteristics, such as a fingerprint, face scan, or voice analysis, protected under Article 9 GDPR<\/td>\n<\/tr>\n
Voice cloning<\/strong><\/td>\nA process in which AI analyzes a person\u2019s voice samples and generates new recordings that sound like that person<\/td>\n<\/tr>\n
AI Act<\/strong><\/td>\nRegulation (EU) 2024\/1689 on artificial intelligence systems<\/a>, which entered into force on 1 August 2024, with provisions applying in stages through 2 August 2026<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

The \u0141ukomski case: Poland\u2019s first lawsuit over voice cloning<\/h2>\n

Who sued whom, and why?<\/h3>\n

Jaros\u0142aw \u0141ukomski, one of Poland\u2019s best-known voice-over artists, active in the market since the mid-1980s, sued JFC Polska sp. z o.o.<\/strong>, a manufacturer of domestic tanks and wastewater treatment systems.<\/p>\n

The case went to the 22nd Intellectual Property Division of the Warsaw Regional Court<\/strong>. The first hearing took place on 23 January 2026. The next was scheduled for 10 April 2026.<\/p>\n

What is the claimant seeking?<\/h3>\n\n\n\n\n\n\n\n\n\n
Claim<\/th>\nAmount or form<\/th>\nLegal basis<\/th>\n<\/tr>\n<\/thead>\n
Compensation for infringement of personal rights<\/td>\nPLN 50,000<\/td>\nArticle 24 of the Civil Code in conjunction with Article 448<\/td>\n<\/tr>\n
Unjust enrichment, including the saved voice-over fee and advertising effect<\/td>\nPLN 102,000<\/td>\nArticle 405 of the Civil Code<\/td>\n<\/tr>\n
Public apology<\/td>\nForm not disclosed<\/td>\nArticle 24 of the Civil Code<\/td>\n<\/tr>\n
Prohibition on further infringement, including removal of the recording<\/td>\nNot stated<\/td>\nArticle 24 of the Civil Code<\/td>\n<\/tr>\n
Total minimum financial exposure<\/strong><\/td>\nApproximately PLN 152,000<\/strong><\/td>\nExcluding litigation costs, expert fees, and interest<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

What happened?<\/h3>\n

An advertisement for domestic wastewater treatment systems was published on YouTube and in a Facebook campaign<\/a>. Its reach was several hundred thousand people. The voice was not a playback of an existing recording. It was speech synthesized by AI<\/strong> from samples of \u0141ukomski\u2019s voice available online.<\/p>\n

Mikrofonika, a voice bank with which \u0141ukomski works under the VO+AI project, analyzed the recording. AI tools indicated a 98 to 99 percent<\/strong> match between the advertisement and the voice artist\u2019s voice. Sound engineers from leading studios confirmed the result. At the first hearing, the court granted a request to appoint a forensic audio expert.<\/p>\n

One important point: \u0141ukomski has an agreement with Mikrofonika covering lawful cloning of his voice and creation of an AI voice engine. JFC Polska could have used a lawful solution. Instead, an unauthorized deepfake was created.<\/p>\n

Mikrofonika filed a separate lawsuit against JFC Polska, seeking protection of the voice artist\u2019s personal rights and its own interests as a voice bank.<\/p>\n

Three areas of legal risk when AI clones a voice<\/h2>\n

The \u0141ukomski case reveals three independent liability regimes that may overlap. Each creates separate consequences.<\/p>\n

The Civil Code: personal rights and voice likeness<\/h3>\n

Article 23 of the Polish Civil Code does not expressly list voice as a personal right. It provides an open catalogue: health, liberty, honor, name, and image. Polish legal scholarship and case law have long expanded that catalogue to include new rights, including voice.<\/p>\n

How the case law developed:<\/strong><\/p>\n\n\n\n\n\n\n\n
Decision<\/th>\nDate<\/th>\nMain holding<\/th>\n<\/tr>\n<\/thead>\n
Supreme Court, II CK 330\/03<\/td>\n20 May 2004<\/td>\nImage includes facial, bodily, and vocal features if they make it possible to identify a person<\/td>\n<\/tr>\n
Supreme Court, II CSK 207\/07<\/td>\n3 October 2007<\/td>\nIntroduced the concept of a \u201cvoice image\u201d<\/strong>: voice as part of the image protected under Article 23 of the Civil Code, provided it is recognizable<\/td>\n<\/tr>\n
Supreme Court, II CSKP 659\/22<\/td>\n7 March 2023<\/td>\nVoice as an independent personal right<\/strong>, not merely part of image but a separately protected value<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

There is one condition for protection: listeners must be able to recognize the voice<\/strong>. The question is not whether the recording is technically authentic. It is whether listeners associate the voice with a specific person.<\/p>\n

The same condition was confirmed by the Berlin Regional Court in August 2025 in the case of dubbing actor Manfred Lehmann, the German voice of Bruce Willis in films. The court found an infringement of the right to one\u2019s voice using the so-called attribution effect<\/strong> (Zuordnungswirkung): whether listeners get the impression that the voice comes from that person or that the person consented. The technical method used to produce the voice, whether recording, human imitation, or AI synthesis, is legally irrelevant.<\/p>\n

The Berlin court ordered the infringement to cease and awarded monetary relief in the form of a so-called fictional license, valued by reference to market rates for a voice artist\u2019s participation in similar productions.<\/p>\n

What this means for your company:<\/strong> if AI generates a voice that listeners associate with a specific person, regardless of whether you entered their name in the prompt, the risk of infringing personal rights is real. A prompt such as \u201ca voice like [Name]\u201d may be evidence in court of intentional infringement.<\/p>\n

Contact us<\/a> if you want to assess whether your AI-voice marketing materials could create a risk of claims.<\/p>\n

GDPR: voice as biometric data<\/h3>\n

A voice processed by AI systems may constitute biometric data<\/a><\/strong> within the meaning of Article 4(14) GDPR. The Polish Data Protection Authority (UODO) has consistently confirmed this.<\/p>\n

Three classification criteria under the guidance of the European Data Protection Board<\/a> (EDPB):<\/p>\n

    \n
  1. Nature of the data: a person\u2019s physical, physiological, or behavioral characteristics<\/li>\n
  2. Specific technical processing: analysis of the voice by an AI algorithm<\/li>\n
  3. Purpose: unique identification of a person<\/li>\n<\/ol>\n

    A voice recording without biometric analysis is ordinary personal data. It becomes biometric data when it undergoes technical processing aimed at identification, and voice cloning meets this condition.<\/p>\n

    UODO positions confirming this classification:<\/strong><\/p>\n\n\n\n\n\n\n\n\n
    Context<\/th>\nUODO position<\/th>\n<\/tr>\n<\/thead>\n
    Electronic medical registration with a voice assistant<\/td>\nA system with a voice assistant processes \u201cdata relating to biometric characteristics\u201d<\/td>\n<\/tr>\n
    Voicebots for booking medical tests<\/td>\n\u201cAnalysis of a unique biometric characteristic such as the human voice makes it possible to identify a person\u201d<\/td>\n<\/tr>\n
    Recording conversations in public offices<\/td>\nNo legal basis to record audio, an infringement of the data-minimization principle<\/td>\n<\/tr>\n
    Warsaw Sobering-Up Station<\/td>\nA PLN 10,000 fine for recording audio as part of surveillance without a legal basis<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    Consequences for companies cloning voices:<\/strong><\/p>\n

      \n
    1. Voice cloning is biometric-data processing, triggering the prohibition in Article 9(1) GDPR.<\/li>\n
    2. Processing requires an explicit legal basis. For commercial use, this will most often be the explicit consent of the person whose voice is involved.<\/li>\n
    3. A DPIA (data protection impact assessment)<\/a> is required. UODO indicates that biometric-data processing generally requires one.<\/li>\n
    4. Fines: up to EUR 20 million or 4% of annual turnover<\/a><\/strong>, whichever is higher.<\/li>\n<\/ol>\n

      The President of UODO, Miros\u0142aw Wr\u00f3blewski, has expressly stated that online platform operators are responsible for content containing another person\u2019s voice and must verify whether the person publishing it has the consent of the person whose data is involved.<\/p>\n

      Criminal Code: personal liability for the management board<\/h3>\n

      The third area is criminal liability. Two provisions may apply:<\/p>\n\n\n\n\n\n\n
      Provision<\/th>\nConduct<\/th>\nPenalty<\/th>\n<\/tr>\n<\/thead>\n
      Article 190a(2) of the Criminal Code<\/td>\nImpersonating another person and using their image, personal data, or other information that enables identification<\/td>\nUp to 3 years\u2019 imprisonment<\/td>\n<\/tr>\n
      Article 286 of the Criminal Code<\/td>\nFraud, if an advertisement using a fake voice suggests that a specific person endorses a product<\/td>\nUp to 8 years\u2019 imprisonment<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

      An advertisement using a deepfake voice to suggest that a well-known person endorses a product, even though they never consented, may meet the elements of both offenses.<\/p>\n

      Criminal liability is not covered by liability insurance.<\/strong> It is a personal risk for board members<\/a> who approved a campaign using a cloned voice.<\/p>\n

      Recital 133 of the AI Act expressly identifies \u201cimpersonation of other persons\u201d and \u201cfraud\u201d as new types of risk generated by AI systems creating synthetic content, including audio. The EU legislator recognizes audio deepfakes as a risk requiring regulation.<\/p>\n

      The AI Act: another layer of duties from 2026<\/h2>\n

      The AI Act<\/a> (Regulation (EU) 2024\/1689) introduces duties that directly concern companies using voice generators:<\/p>\n

        \n
      1. Marking synthetic content (Article 50(2)):<\/strong> providers of AI systems generating synthetic audio must ensure outputs are marked in a machine-readable format and detectable as artificially generated.<\/li>\n
      2. Disclosure of deepfakes (Article 50(4)):<\/strong> deployers of AI that generates deepfake audio must disclose that the content was artificially generated or manipulated.<\/li>\n
      3. Penalties:<\/strong> up to EUR 35 million or 7% of global turnover<\/strong> for prohibited AI practices; up to EUR 15 million or 3% of turnover<\/strong> for breaching obligations concerning high-risk systems<\/a>.<\/li>\n<\/ol>\n

        The disclosure duty rests with the company using the AI-generated voice in an advertisement<\/strong>, not only the tool provider. Even if you buy a voice-over from an agency, you are responsible for how the content is used as the \u201cdeployer\u201d under the AI Act.<\/p>\n

        An exception applies to content with an obviously artistic, satirical, or fictional character, but a disclosure duty still exists, in a limited form.<\/p>\n

        The Polish AI Systems Act<\/strong>, a draft adopted on 31 March 2026, establishes a supervisory authority, the Commission for the Development of AI Safety (KRiBSI). Its members will include, among others, the President of UOKiK and representatives of KRRiT, PIP, KNF, and the President of UKE. This body will enforce AI Act duties in Poland.<\/p>\n

        Cumulative regimes: risks do not cancel each other out<\/h3>\n

        Recital 63 of the AI Act expressly states that classifying an AI system \u201cshould not be interpreted as indicating that the use of that system is lawful under other Union or national law.\u201d A company may breach the AI Act, GDPR, and the Civil Code at the same time, and sanctions can accumulate.<\/p>\n\n\n\n\n\n\n\n\n\n
        Legal regime<\/th>\nMaximum financial penalty<\/th>\nPersonal liability for board members<\/th>\n<\/tr>\n<\/thead>\n
        Civil Code (personal rights)<\/td>\nNo limit (depends on the claim)<\/td>\nNo<\/td>\n<\/tr>\n
        GDPR (biometric data)<\/td>\nEUR 20 million or 4% of turnover<\/td>\nYes, in certain cases<\/td>\n<\/tr>\n
        Criminal Code (impersonation)<\/td>\nNot applicable<\/td>\nYes, up to 3 years\u2019 imprisonment<\/td>\n<\/tr>\n
        Criminal Code (fraud)<\/td>\nNot applicable<\/td>\nYes, up to 8 years\u2019 imprisonment<\/td>\n<\/tr>\n
        AI Act (transparency duties)<\/td>\nEUR 35 million or 7% of turnover<\/td>\nYes, in certain cases<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

        AI platform terms do not protect your company<\/h2>\n

        Voice-generation platforms, including ElevenLabs, Play.ht, and Murf, shift all responsibility for third-party-rights infringements to the user in their terms.<\/p>\n

        What the terms say:<\/strong><\/p>\n\n\n\n\n\n\n\n
        Platform<\/th>\nLiability model<\/th>\n<\/tr>\n<\/thead>\n
        ElevenLabs<\/td>\nAn indemnity clause: the user releases the platform from liability for IP, privacy, and model-generated content claims; users are prohibited from infringing third-party rights<\/td>\n<\/tr>\n
        Play.ht<\/td>\nFree plan: no right to commercial use; paid plans: a commercial license, but no guarantee concerning third-party rights<\/td>\n<\/tr>\n
        Murf<\/td>\nThe user owns the content; the platform receives a non-exclusive license; no visible clauses protecting against third-party claims<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

        Limits on the effectiveness of these clauses under Polish law:<\/strong><\/p>\n

          \n
        1. Contract terms cannot exclude liability for harm caused intentionally.<\/li>\n
        2. Terms of service do not exclude tort liability to injured third parties.<\/li>\n
        3. Mandatory GDPR and AI Act provisions cannot be waived by terms of service.<\/li>\n
        4. In consumer relationships, terms that seriously infringe consumers\u2019 interests may be found unfair.<\/li>\n<\/ol>\n

          Buying a voice license from an AI platform does not protect you against a claim from the person whose voice was cloned. The license covers use of the tool, not the right to someone else\u2019s voice.<\/p>\n

          What to do in the next 48 hours: a checklist for CEOs and CMOs<\/h2>\n

          Immediate actions (priority: you must do this)<\/h3>\n
            \n
          1. Audit marketing materials.<\/strong> Review every commercial, podcast, onboarding material, voice chatbot, and IVR system using a synthetic voice. Identify anything that may imitate a specific person\u2019s voice, even if their name was not included in the prompt.<\/li>\n
          2. Review prompts.<\/strong> Check which prompts were used to generate AI voices. A prompt such as \u201ca voice like [Name]\u201d or \u201ca voice resembling [Person]\u201d may be evidence of intentional infringement of personal rights.<\/li>\n
          3. Legal sign-off before publication.<\/strong> Require legal review of every advertising material with AI voice-over before publication. Include in the checklist the voice source, the prompt used to generate it, and the platform license.<\/li>\n
          4. Update agency agreements.<\/strong> Add clauses to agreements with advertising agencies and subcontractors addressing liability for voice-right infringements (voice likeness) and representations that AI materials were used lawfully.<\/li>\n<\/ol>\n

            Actions within 2 to 4 weeks (priority: you should do this)<\/h3>\n
              \n
            1. Internal AI-voice policy.<\/strong> Develop rules for using AI voice-generation tools, including a ban on prompts referring to specific people and a duty to document the voice model\u2019s source.<\/li>\n
            2. Team training.<\/strong> Train marketing, creative, and procurement teams on legal risks, with particular focus on voice recognizability and criminal consequences.<\/li>\n
            3. Review AI platform licenses.<\/strong> Analyze the terms of voice-generation platforms for third-party-rights liability and the scope of indemnification.<\/li>\n<\/ol>\n

              Actions worth considering<\/h3>\n
                \n
              1. A \u201cvoice clearance\u201d procedure.<\/strong> Consider a process similar to music clearance: before using an AI voice, check whether it is recognizable as a specific individual. Involve an independent listener or test panel.<\/li>\n<\/ol>\n

                Who is liable: the company or the agency?<\/h2>\n

                This is one of the most common questions. The answer: primarily the company that commissioned the advertising material<\/strong>.<\/p>\n

                The agency produced the commercial, but your company published and approved it and benefited from it. In the \u0141ukomski case, JFC Polska is the defendant, as the company that commissioned and published the advertisement, not the agency that produced it.<\/p>\n

                Of course, the company may pursue recourse claims against the agency, but only if the contract provides for them. Hence the recommendation: update subcontractor agreements with clauses transferring liability for voice-right infringements.<\/p>\n

                The AI Act confirms this principle: responsibility for AI rests not only with the provider, the tool\u2019s developer, but also with the deployer of the system<\/strong>. A company using a ready-made AI tool is responsible for how it uses it.<\/p>\n

                Not knowing, \u201cwe did not know this was AI,\u201d may not be enough as a defense. A marketing director who approves advertising material without checking the voice source exposes both the company and themselves.<\/p>\n

                How we can help<\/h2>\n

                The \u0141ukomski case is a precedent that will shape Polish case law on protection of voice from AI. Whatever the outcome, its existence alone changes the legal environment for every company using synthetic voices in marketing communications.<\/p>\n

                We support technology and e-commerce companies<\/a> in organizing their use of AI, from auditing marketing materials and preparing internal policies to updating agency and subcontractor agreements.<\/p>\n

                Our support includes:<\/p>\n

                  \n
                1. Audit of marketing materials<\/strong> for AI-voice risks, identifying recordings that may infringe the rights of specific people.<\/li>\n
                2. Preparing an internal policy<\/strong> for AI voice generators, including prompting rules, source documentation, and approval procedures.<\/li>\n
                3. Updating agreements<\/strong> with advertising agencies and subcontractors with clauses protecting against claims related to audio deepfakes.<\/li>\n
                4. GDPR and AI Act compliance analysis<\/strong> to check whether voice processing meets requirements for biometric data and transparency duties.<\/li>\n
                5. Training for marketing and creative teams<\/strong> on legal risks, procedures, and responsibility.<\/li>\n<\/ol>\n

                  We combine IP law, data protection, and AI regulation in one coherent model tailored to how your company actually uses technology.<\/p>\n

                  AI voice in advertising: organize the rules before a court does it for you<\/h2>\n

                  Polish law protects voice as a personal right. GDPR classifies it as potentially biometric data. The AI Act imposes transparency duties on every company using synthetic audio content. The Criminal Code provides for personal liability for board members for impersonating others.<\/p>\n

                  The \u0141ukomski case shows that these risks are not theoretical. The lawsuit has been filed, an expert has been appointed, and the next hearing is scheduled.<\/p>\n

                  If your company uses AI-generated voice in marketing or a product, check whether you are doing so lawfully before an inspection, lawsuit, or media report forces you to act under pressure.<\/p>\n

                  Contact us<\/a>. We can help you organize the rules for using AI voice and reduce legal risk.<\/p>\n

                  Frequently asked questions<\/h2>\n

                  Can I safely use an AI voice generator in advertising if I do not name a specific person in the prompt?<\/strong><\/p>\n

                  Not necessarily. Omitting a name from the prompt does not eliminate the risk. What matters is whether listeners recognize the voice, not how you generated it. If listeners associate the voice with a specific person, an infringement of personal rights may be found regardless of the prompt. The Berlin court judgment from August 2025 confirmed this principle: the technical method used to produce the voice is legally irrelevant.<\/p>\n

                  Who is liable: my company or the advertising agency?<\/strong><\/p>\n

                  Primarily the company that published the material and benefited from it. In the \u0141ukomski case, the company that commissioned the advertisement, not the production agency, is the defendant. You may pursue recourse against the agency, but only if the contract provides for it. That is why it is worth updating agreements with clauses transferring liability for voice-right infringements.<\/p>\n

                  Does buying a voice license from an AI platform, such as ElevenLabs or Play.ht, protect me legally?<\/strong><\/p>\n

                  A license covers use of the tool, not the right to someone else\u2019s voice. Platforms shift responsibility for third-party-rights infringements to users in their terms. Indemnity clauses do not exclude your tort liability to the person whose voice was cloned or waive duties under GDPR and the AI Act.<\/p>\n

                  Could board members actually face criminal liability for approving an AI-voice campaign?<\/strong><\/p>\n

                  Yes. Article 190a(2) of the Criminal Code, impersonating another person, carries a penalty of up to 3 years\u2019 imprisonment, while Article 286, fraud, carries up to 8 years. If a deepfake-voice advertisement suggests that a well-known person endorses a product without their consent, the elements of these offenses may be met. Criminal liability is not covered by liability insurance.<\/p>\n

                  What amounts could my company face in claims?<\/strong><\/p>\n

                  In the \u0141ukomski case, the total minimum exposure is approximately PLN 152,000, excluding litigation costs and interest. Potential GDPR fines of up to EUR 20 million or 4% of turnover and AI Act fines of up to EUR 35 million or 7% of turnover may also apply. The risks can accumulate: a company may face claims on several fronts at once.<\/p>\n

                  Will the \u0141ukomski judgment affect other companies?<\/strong><\/p>\n

                  Yes. As the first lawsuit of its kind in Poland, the Warsaw Regional Court\u2019s decision will shape Polish courts\u2019 practice in AI voice-cloning cases. Regardless of the outcome, the proceeding signals the direction in which case law is developing. Companies using synthetic voices should adapt their processes now rather than wait for a final judgment.<\/p>\n","protected":false},"excerpt":{"rendered":"

                  A Polish voice-cloning lawsuit shows how AI-generated speech can expose marketers to personal-rights, data-protection and criminal claims.<\/p>\n","protected":false},"author":9,"featured_media":1906,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":"","sip_alt_url":"","sip_en_pl_url":"","sip_pair_uuid":"","sip_pair_state":""},"categories":[1577],"tags":[1545,1537,1543,1632],"specialization":[1211],"practice_area":[],"class_list":["post-3910","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-it-ai-ip-law","tag-ai-act-en","tag-intellectual-property-en","tag-it-ai-and-ip-law","tag-systemy-wysokiego-ryzyka","specialization-it-ai-ip-law"],"acf":[],"_links":{"self":[{"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/posts\/3910","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/users\/9"}],"replies":[{"embeddable":true,"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/comments?post=3910"}],"version-history":[{"count":1,"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/posts\/3910\/revisions"}],"predecessor-version":[{"id":4161,"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/posts\/3910\/revisions\/4161"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/media\/1906"}],"wp:attachment":[{"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/media?parent=3910"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/categories?post=3910"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/tags?post=3910"},{"taxonomy":"specialization","embeddable":true,"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/specialization?post=3910"},{"taxonomy":"practice_area","embeddable":true,"href":"https:\/\/sawaryn.com\/en\/wp-json\/wp\/v2\/practice_area?post=3910"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}