AI face generation includes two related but distinct capabilities: generating entirely new synthetic faces (as seen in 'This Person Does Not Exist') and producing realistic faces based on reference images of real individuals. The second capability, identity-preserving face generation, is most relevant to privacy concerns. It allows for the creation of intimate images of real individuals using their reference photos from social media. While these images are entirely AI-generated, they depict a real identifiable person and may raise privacy risks.

Key facts about this term

  1. Identity-preserving generation poses key privacy risks Generic face generation creates fictional people. Identity-preserving face generation creates realistic depictions of real individuals, this is what raises privacy concerns.
  2. Reference photos from social media can suffice Modern identity-preserving face generation tools require as few as 3-5 facial reference photos, all of which may be sourced from a public social media profile.
  3. AI-generated images depicting real individuals raise privacy concerns An AI-generated intimate image that clearly depicts a real, identifiable individual, even if entirely synthetic, potentially raises privacy risks.

Frequently asked questions

Is it legal for AI tools to use my public photos to generate intimate imagery?

No. Generating intimate imagery of an identifiable real person without consent using their public photos raises privacy concerns and may violate federal law.

Can I find AI-generated images of myself that I don't know about?

Yes. ScanErase's biometric scan searches for your face embedding across the web and major platforms, finding AI-generated images that would not appear in a keyword search for your name.