A deepfake is synthetic media (typically video, audio, or both) that depicts a real, identifiable person doing or saying something they did not actually do or say — produced without that person's consent.
The term has been overloaded in popular usage, but the load-bearing word is consent. Same underlying technology (face generation, voice cloning, lipsync) can produce two products with opposite legal and ethical status depending on whether the depicted person agreed to the depiction.
| Consensual cloning | Deepfake | |
|---|---|---|
| Source person | Client / commissioned subject | Target without consent |
| Approval | Required per output | None |
| Legal status | Generally permitted with contract | Often illegal (deepfake laws in CA, TX, IL, EU) |
| Typical use | Done-for-you content, dubbing, accessibility | Disinformation, fraud, non-consensual intimate imagery |
At DFY Content, every clone is built with explicit signed consent from the client. Every output requires the client's approval before publishing. The client owns their clone, scripts, and produced media. The same technical pipeline could in principle be used to deepfake someone — what prevents it is the contract, the workflow, and the deliberate refusal to take clients who can't prove they're the person being cloned.
Common deepfake categories (each typically illegal in major jurisdictions):
- Non-consensual intimate imagery
- Political disinformation depicting real candidates
- Financial fraud using cloned executive voices
- Identity theft via voice or face impersonation