AI Disclosure

Last updated 26 August 2026

Seen & Amplified™ is an AI platform, and you are trusting it with the most personal thing in your business: how you sound. This page says exactly what that involves, in the detail you would want before handing it over.

01Which models we use

We use large language models from Anthropic and Google through API services. Which one handles a given task depends on what the task needs, and we may change or add providers as better options appear. Any provider we add is listed in the sub-processor table in our Privacy Policy before it goes live.

We do not build or train our own models.

02What happens when you generate something

When you press generate, the relevant parts of your profile and content — your voice settings, brand details, and whatever the tool needs for that job — are sent to a provider's API. The model returns a draft. That draft is stored in your account and shown to you.

• Only what the task needs is sent, not your whole account.
• The request is made to produce your output and nothing else.
• FeiModern does not use this traffic to train models. Anthropic does not use commercial API inputs or outputs for model training by default. Google's no-training terms apply to Gemini Paid Services connected to an active Cloud Billing account.

03What Voice Calibration actually does

When you edit a draft before publishing, the platform keeps a pair: what was generated, and what you changed it to. When you publish one untouched, it keeps that too, as a confirmed example.

Those pairs are stored against your account only. They are read back into your own future generations so drafts land closer to how you write. FeiModern does not use them to train a model, they are not pooled across accounts, and no other member's drafts are influenced by them. This is the difference between remembering preferences inside your workspace and training a shared model — the first is what happens here.

04The human stays in the loop

This platform does not act on its own. It does not publish, send, or contact anyone without you choosing it.

• Every generated draft is yours to read, edit, or throw away.
• Scheduling is something you set. Fei can suggest, but you approve.
• Drafted messages to your contacts are drafts until you send them.

The platform is designed around the transparency principles reflected in Article 50 of the EU AI Act: you know when you are working with an AI tool, and you see the output before anyone else does. This description is not a claim that every Article 50 obligation applies to every use of the platform.

05What AI gets wrong

Being straight about the limits is part of the disclosure.

• It can be confidently wrong. Facts, figures, dates and quotes need checking before you publish them.
• It does not know what it does not know. It works from what you gave it. Thin input produces generic output.
• It can land near something that already exists. Phrasings recur. If a line matters, check it is yours.
• It is not advice. Nothing it produces is legal, financial, medical or therapeutic guidance, whatever it sounds like.

You are responsible for what you publish under your own name. That is also the point of the product: it gets you to a draft, and your judgement makes it worth reading.

06Do you have to tell people your content is AI-assisted?

That is your call, and it depends on where you are and where you publish. Some platforms have their own labelling rules, and some jurisdictions are introducing them. Our view is simpler: the content is written from your positioning, your stories and your voice, and you approved every word before it went out. It is your work either way.

07Your data, plainly

08Questions

If something here matters to your decision and this page did not answer it, email hello@feimodern.com and ask. A straight answer before you join is better for both of us than a surprise afterwards.