HeyGen and GDPR / AI Act: avatars, voice cloning and anti-impersonation

HeyGen and GDPR / AI Act: avatars, voice cloning and anti-impersonation

In brief

  • HeyGen (US publisher): realistic avatars, voice cloning, synchronised video translation. Acute identity issues: written and specific consent to image and voice (cloning), transfers outside the EU (CLOUD Act), marking/disclosure (Article 50). Realism creates an impersonation risk (executive fraud): impose dual validation of video/voice orders. For sensitive uses, evaluate a European alternative.

Overview and data processed

HeyGen is a US-based AI video platform: realistic avatars, voice cloning and video translation with lip-syncing. It allows you to create an avatar and voice from a real person.

HeyGen processes the image and voice of cloned individuals (potentially biometric data), text and generated videos. US publisher: exposure to the CLOUD Act.

Risks for your organisation

Voice cloning and realistic avatars make consent absolutely central: an unconsented voice clone is a direct infringement. In addition, there are transfers outside the EU and the transparency obligation.

Main GDPR risks:

  • Reproducing face/voice requires consent (right to image, voice as personal data).
  • Custom avatar/clone: written, specific, revocable consent; purposes.
  • Transfer outside the EU (CLOUD Act); SCCs to be put in place.
  • Origin of training data (scraping prohibited).

GDPR obligations

  1. Obtain written and specific consent before any cloning (image + voice).
  2. Frame uses and revocation; inform individuals.
  3. Sign the DPA; frame transfers (SCCs).
  4. Enter in the record of processing activities; DPIA if large-scale biometrics.
  5. Involve the DPO; train teams. No DPO? See our outsourced DPO.

AI Act obligations

Article 50 (applicable 2 August 2026, technical deadline 2 December 2026): the provider marks technically; the deployer who produces a deepfake must disclose clearly. Lighter regime for artistic/satirical content. GPAI; AI literacy. Prohibition (Article 5): untargeted facial scraping prohibited. See the AI Act guide.

AI Act risks: generated content not marked/not disclosed (Article 50); deepfakes (real face/voice) without disclosure or consent; data from scraping (to be verified).

Cybersecurity and good practices

Cyber risks: fraud by vocal/video cloning (executive impersonation), transfer outside the EU, access to clones to be protected; NIS 2 depending on size.

Good practices: written and specific consent for any clone (register); systematic marking/disclosure; frame transfers (SCCs); anti-impersonation procedure (dual validation of orders); AI charter, training; sovereign alternative for sensitive matters.

Checklist:

  • Written consent (image + voice) before cloning, uses and revocation framed, individuals informed
  • Marking/disclosure (Article 50), DPA + transfers framed (SCCs), data origin verified
  • Anti-impersonation procedure in place, DPIA if large-scale biometrics, teams trained

Sanctions, conclusion and recommendations

Sanctions: GDPR (up to €20M or 4%); AI Act (€15M or 3% for Article 50, €35M or 7% for prohibited practices); civil and criminal — right to image, articles 226-8 and 226-8-1 of the French Criminal Code.

Recommendation: the realism of HeyGen's avatars and voices creates a major impersonation risk. Systematic written consent, marking/disclosure, framed transfers and anti-fraud procedure are essential; for sensitive uses, evaluate a European alternative. Compare with Synthesia and Sora. See the AI tools comparison.

Take action

Assess the GDPR and AI Act risk level of your HeyGen uses with the Data Comply One self-assessment.

Discover the AI Act software →Test my compliance for free
Compare this tool with the 70 AI analysed — hosting, training, DPA and verdict.See the full comparison →

Frequently asked questions

Is HeyGen GDPR-compliant?

It can be provided you obtain consent for cloning (image and voice), frame transfers (DPA, SCCs) and mark/disclose generated content.

Does voice cloning require consent?

Yes: voice is personal data; cloning it requires written, specific and revocable consent.

Can HeyGen be used for fraud?

The realism of avatars and voices creates an impersonation risk; put in place dual validation for any order transmitted by video or voice message.

Must HeyGen videos be flagged as AI-generated?

Yes: Article 50 requires marking and, for a deepfake, disclosure. A mention in the T&Cs is insufficient.