AI can now make a person appear to have said something they never said. Their voice can be imitated. Their face can be placed into a photo or video that was never filmed. And at worst, their identity can be used to create sexual material without any consent whatsoever.
On 9 September 2026, Yle published an article about the spread of deepfakes in Finland. It tells of women and girls whose real photos have been used to produce AI-generated sexual material without their permission. According to Yle, the phenomenon has surfaced in schools, workplaces and the Finnish Defence Forces, among other places.
If AI uses a person's face, voice or identity, they must have the right to decide about it.
This conversation matters to us, too. At Persona Avatar, we build official AI avatars in collaboration with our Talents. A Persona Avatar representing a real person is never made of just anyone, or without that person's involvement. Every such AI avatar is based on an agreed, contractual partnership between Persona and the Talent — and that is exactly why we also need to talk about where the limits of this technology lie.
Deepfakes and AI avatars are easily mentioned in the same breath. Both can use generative AI and synthetic image, video and audio — but that does not make them the same thing. The crucial difference lies in who gave permission, what the technology is used for, and whether the audience knows what it is looking at.
In the cases highlighted by Yle, people's real photos have been used to produce sexual material without their consent. The person in the image has not chosen to participate and has no control over what is done with their identity. According to Yle's investigation, the phenomenon reaches from well-known Finns to ordinary people and minors.
At Persona Avatar, the starting point is the opposite. The Talent knows their AI avatar is being built. They give permission, take part in the process and sign an agreement with Persona on how their identity and materials may be used. For us, consent is not a small legal detail surrounding the technology — it is the foundation of the entire model.
We have also received feedback that our AI avatars do not always look like entirely real people. And that is an interesting observation, because we have grown used to judging synthetic video above all by how well it manages to fool the eye.
That can be precisely the purpose of a deepfake: to make the viewer believe they are seeing a real image, video or person. The harder the manipulation is to detect, the more convincing it can be. A Persona Avatar has a different job: the purpose of an AI avatar is not to make you forget you are talking with AI, but to enable the conversation. The user should know at all times which is the real person and which is the AI avatar built in collaboration with them.
That is why a good AI avatar cannot be evaluated only by asking: “Does this look exactly like a real person?” Far more important questions are: does the real person know about this? Have they given their permission? Does the user know they are talking with AI? And who decides what this digital identity may be used for?
Technical possibility is not the same as permission.
The cases brought to light by Yle show concretely what happens when this line disappears.
Finding someone's face on Instagram is not consent to use it in AI-generated material. Becoming a public figure is not consent to creating a digital copy of them. Publishing a video online does not mean the voice in it is free to use.
This principle becomes ever more important as the technology advances. Creating a digital version of a person is constantly becoming easier and more convincing. In Yle's story, one producer of deepfake content even boasted about being able to generate up to 200 pornographic deepfake images a day.
That is exactly why the question should shift from whether we can do this to whether we have permission to do this.
This idea is the starting point of Persona Avatar. An official AI avatar is not built by scraping a public figure's photos, videos and speech from the internet and turning them into a digital version of that person.
Persona works with the Talent on a contractual basis. The Talent gives permission to use their face, voice and agreed materials, and participates in building their own AI avatar. The avatar is given a knowledge base and clear limits on what and how it can discuss. The Talent might be, for example, an athlete, an expert, a mentor, a content creator or an election candidate. What they all share is that they have chosen to take part themselves. We do not make AI avatars of people without their knowledge — we build them together with them.
In a responsible AI avatar, transparency works in both directions: the Talent must know how their identity is used, and the user in turn must know who — or rather what — they are talking with.
When you open the AI avatar of an athlete, a mentor or an election candidate in Persona Avatar, the point is not to make you believe that the real person is at the other end talking with you in real time. You are talking with an AI avatar.
Transparency does not make the technology any less interesting. Quite the opposite. When we do not need to pretend the avatar is something it is not, we can focus on what entirely new things can be done with it.
Yle's news story is a reminder of how serious the consequences of misusing synthetic media can be. Journalist Johanna Vehkoo, interviewed by Yle, describes the phenomenon in Finland as nothing less than an epidemic and warns against its normalisation. That is why deepfakes must be talked about, abuse must be addressed, and the people targeted must be protected.
At the same time, we need more precise language about what is being built with AI. Not every AI-generated video is a deepfake. Not every digital avatar is an attempt to pose as another person. Technology can enable both serious abuse and entirely new, consent-based ways of communicating.
The difference is not necessarily visible in the pixels first, but in how and why the content came to be. Is the real person involved? Have they given permission? Has the collaboration been agreed? Is it clear to the user that this is an AI avatar? Have limits been defined for the avatar? These are the questions we should be talking about more as AI becomes an ever bigger part of our everyday lives.
Faces and voices are not free raw material. An official AI avatar begins with a person who has chosen to take part.
In the end, the deepfake discussion is not only about what AI can do. It is also a conversation about ownership, responsibility and consent.
At Persona Avatar, AI avatars representing real people are not made of just anyone. An official AI avatar is born from a collaboration between Persona and the Talent, in which the person themselves has decided to bring their digital version to Persona Avatar.
Technology will keep evolving. Digital people will become ever more convincing, and telling real and synthetic content apart may become harder. Then the most important question for us is not how real a person AI can create, but: with whose consent is it created?
