|Artguilty

the rules behind our rendered spokesperson

Signe fronts artguilty's short videos about the systems behind online stores. She is rendered. We made her from prompts, reference images and one spec file, and no real person's photo was used for her face. Her voice is a stock voice from our video tool's library; we cloned no one.

Before her first video went out, we wrote down the rules she follows and checked them against the law that reaches a business advertising with a made character. This is what we found, as of November 2026. It is our reading for our own use. It is not legal advice; for your own situation, ask a lawyer.

is it legal to use an AI spokesperson?

Yes, in the US and in Europe, with conditions. No rule we found bans a made character. What the rules police is deception:

  • a viewer misled about what the character is
  • a claim of experience nobody had, such as a review or a testimonial
  • a real person's face or voice used without permission

what US advertising law asks

The Federal Trade Commission Act bans deceptive advertising. Two sets of rules under it reach a spokesperson like Signe.

  • The Endorsement Guides, revised in 2023, count virtual influencers as endorsers. An endorsement has to reflect real opinions and real use, so a character can't say she tried a product or liked a service.
  • The rule on fake reviews and testimonials, in force since October 2024, bans reviews and testimonials from someone who doesn't exist, and names AI-generated ones. Knowing violations can bring civil penalties.

Signe speaks for us, about our own work, so she may count as the brand speaking rather than as an endorser. We don't lean on that. She never endorses anything, never claims to have used a product, and never appears as a client, a customer or a review.

whose face, whose voice

Most US states protect a person's name, face and voice from being used to sell without permission, under what is called the right of publicity. A character made from a real person's photo, or speaking in a cloned voice, could be using that person without permission. Signe's face came from prompts, and we never ask an image model to make her look like someone.

New Jersey, where artguilty is registered, passed a deepfake law in April 2025. It covers media that realistically shows a person saying or doing something they never did, made or shared to commit a crime. As we read it, a character who is nobody, in a labeled video, is outside what it describes.

what Europe asks

We work for clients in Europe too. Since August 2, 2026, Article 50 of the European Union's AI Act requires anyone who uses AI to make a deep fake to disclose that the content was artificially generated. The duty is lighter for work that is evidently fictional or creative: the disclosure can be made in a way that doesn't spoil the work. The Act defines a deep fake as content resembling existing people, places or events that would falsely appear real. Whether a character who resembles no one counts is open. We label her anyway.

what the platforms ask

  • YouTube requires creators to disclose AI-generated content that seems realistic, through the "AI use" setting at upload. Creators who keep skipping it can have a label added for them, videos removed, or their Partner Program membership suspended.
  • TikTok requires creators to label realistic AI-generated content, and labels it automatically when the file carries Content Credentials.
  • Instagram, Facebook and Threads may require an AI label on photorealistic video and realistic audio, and may penalize content shared without one.

We switch the label on wherever a platform has one. Where it has none, the caption ends with the same line as the video.

the rules we wrote for her

  1. She is never a team member, a client, a customer or a testimonial.
  2. She never answers the question of what she is, and we never dodge it. Every video ends on the same card: "signe is rendered. the advice isn't."
  3. Anyone who asks us sincerely gets a straight answer and a link to her page.
  4. Prices, offers and requests to book come in our voice, in the caption or on the card, never from her.
  5. She never messages anyone one to one. A direct message from a character the reader may not know is rendered is a different case from a labeled public video, and we have no rule for it yet.
  6. No real person's face or voice, ever.
  7. The advice is the team's real work. She does the talking.

what is still open

  • Whether a brand's own character counts as an endorser or as the brand speaking. Our rules hold either way.
  • Whether Europe's deep fake label covers a character who resembles no one.
  • State AI laws keep arriving, most of them aimed at election deepfakes, impersonation of real people, and chatbots that hide what they are. We check again before anything changes about how she's used.

before your business uses one

  • Write down what the character will never be: a customer, a reviewer, an employee.
  • Decide where the disclosure sits on each channel: the platform label, a closing card, the caption, and one page that answers the question straight.
  • Keep a written record of where the face and voice came from: the tools, their terms, and that no one was copied.
  • Keep prices and offers in the brand's voice.
  • Decide the answer to "are you real?" before the first comment asks it.
  • Check the rules of every market you sell into, Europe's included.

the same work, for your business

Signe is how we use AI on our own channels. Deciding which tasks to hand to AI, and saying which ones it isn't good at, is our AI adoption work. It starts with a stack review: one week, fixed fee, from $1,500, with written findings and a prioritized list that is yours either way.

sources