AI Transparency
Last updated: 30 September 2026
feega is built on AI. This page explains, in the spirit of Regulation (EU) 2024/1689 (the “AI Act”), which AI features exist, which models power them, how outputs are marked, how requests are moderated and what the limits are. It complements the Terms of Service (§7–§10), the Acceptable Use Policy and the Privacy Policy.
1. Where AI is used
Chat agent
What the AI does: a conversational AI assistant in the sidebar that reads and edits your canvas, creates nodes and runs generations on your instruction
External agents (MCP, API, CLI)
What the AI does: AI agents you connect act on your workspace with your API key; feega does not control those agents
Text generation
What the AI does: writes and rewrites captions, scripts, articles and other text nodes
Image and video generation
What the AI does: creates and edits images and video from prompts and reference media
Audio generation
What the AI does: voice-over, music and sound effects
Brand analysis
What the AI does: reads a brand’s public website to suggest voice, palette and assets
Research
What the AI does: searches the web and public social posts to inform content
Moderation
What the AI does: screens every generation prompt before it reaches a model (§5)
You always know you are talking to an AI. The chat agent is an AI system, not a person. Nothing in the Service is presented as a human reply.
2. Models and providers
Text, image, video and audio are generated by third-party AI model providers, accessed through platforms such as OpenRouter. The catalogue changes as models are released and withdrawn.
The model used is shown in the Service before you run a generation and recorded in your usage log. For each medium the Service suggests a few recommended models; you can pick any other available model.
Uncensored models are offered only inside NSFW mode, which is not yet available.
Moderation uses an automated classifier and an AI reviewer (§5).
The providers we use as processors are listed in Sub-processors. feega does not train models, and does not use your content to train its own or third-party models.
3. Labelling of generated content
Machine-readable marking. Images and videos generated by the Service carry an IPTC DigitalSourceType marker in embedded XMP metadata: trainedAlgorithmicMedia for fully generated media, compositeWithTrainedAlgorithmicMedia for media that combine your material with generated elements. The marker is inserted without re-encoding the file. If marking fails for a technical reason, the file is still delivered unmarked rather than lost.
Signed Content Credentials (C2PA) and visible “AI-generated” labels across the Service are being rolled out. In NSFW mode every result will be labelled as AI-generated, including in the downloaded file name.
Your duty. Some platforms strip metadata on upload. When you publish a deepfake of real people, places or events, or AI text on matters of public interest without human editorial review, you must disclose that it is AI-generated (Terms §10).
4. Deepfakes and real people
Sexual or nude content depicting a real, identifiable person is always refused.
Deepfakes made to deceive, defame or harass are prohibited in every mode.
Using a real person’s likeness or voice requires their consent; do not present synthetic people or voices as real customers, employees or testimonials.
In NSFW mode an additional check refuses prompts and references that identify a real person.
5. Automated moderation
Every generation prompt, including reference descriptions, is screened before it reaches a model:
Classifier. An automated classifier scores the prompt against a fixed list of categories.
AI review. When the classifier is in doubt, an AI reviewer decides.
Fail-closed. If the review cannot run, the request is refused, not let through. In NSFW mode a classifier outage also refuses the request.
Sexual content involving minors or apparent minors
Standard mode: refused
NSFW mode: refused
Sexual content of a real, identifiable person
Standard mode: refused
NSFW mode: refused
Non-consensual sexual content
Standard mode: refused
NSFW mode: refused
Graphic violence and gore
Standard mode: refused
NSFW mode: refused
Sexual content involving animals
Standard mode: refused
NSFW mode: refused
Self-harm
Standard mode: refused
NSFW mode: refused
Hate
Standard mode: refused
NSFW mode: refused
Weapons and terrorism
Standard mode: refused
NSFW mode: refused
Any other sexual, erotic or nude content
Standard mode: refused
NSFW mode: allowed for fictional adults
Each decision (stage, verdict, category, scores, reason, acting user) is logged. Moderation is automated and can be wrong in both directions: it refuses a request, it does not judge you.
6. Human oversight and reporting
You decide what to run, what to keep and what to publish. Nothing is published or advertised without an action by you or an agent you authorised.
To have a refusal reviewed by a person, write to support@feega.app.
To report illegal or abusive content, write to support@feega.app (Terms §12, Acceptable Use Policy §5).
7. Limitations
AI outputs can be wrong, invented (“hallucinated”), outdated, biased, offensive, or similar to existing works or other users’ outputs. Text can state false facts with confidence; images and video can contain distorted details, text or anatomy. Review every output before you use, publish or advertise it.
8. Risk classification
feega is a general-purpose content tool. It is not designed or offered for any high-risk use listed in Annex III of the AI Act (for example employment, credit, education, essential services, law enforcement, migration, justice), and such uses are prohibited by the Terms.
The practices prohibited by Article 5 of the AI Act are excluded: feega does not use, and does not allow you to use it for, manipulative or exploitative techniques, social scoring, predictive policing by profiling, untargeted facial-image scraping, emotion recognition at work or school, sensitive biometric categorisation or real-time remote biometric identification.
feega makes no decisions with legal or similarly significant effects on people.
9. Contact
AI questions: support@feega.app · Abuse: support@feega.app · Privacy: privacy@feega.app