Recommender System Guidelines (EU DSA Article 27 Compliance)

Sauci Jenna
Sauci Jenna
  • Updated

Sauci AI uses AI-based recommender systems to enhance user experience by suggesting fictional characters, styles, visual content, or chat themes. All recommendations are generated by artificial intelligence models and do not involve real people, profiling, or advertising data. This page explains how the recommendations work, what factors influence them, and how users can control or modify them in accordance with the EU Digital Services Act (DSA).

1. Purpose of Our Recommender Systems

Our recommendation features exist to:

  • help users discover AI characters that match their interests
  • suggest content categories, visual styles, or themes
  • generate story or roleplay ideas
  • tailor interactions to adult users’ fictional preferences

Recommendations are entirely AI-generated, based on your activity within the platform, and never involve real individuals.

2. What Data Is Used for Recommendations?

Sauci AI uses minimal and non-personal signals to generate recommendations.

We may use:

✔ User-selected preferences

  • Character traits (e.g., personality, appearance)
  • Voice styles
  • Interaction themes
  • NSFW/SFW mode

✔ User behavior within Sauci AI

  • Chats with AI characters
  • Visual styles generated
  • Liked or saved characters
  • Frequently selected prompts

✔ Contextual session data

  • On-site behavior
  • Interaction settings
  • Previously generated content type

We do NOT use:

  • Real-person data
  • Third-party advertising data
  • Social media data
  • Biometric data
  • Demographic profiling
  • Location-based behavioral profiling
  • Cross-app or cross-site tracking

Sauci AI operates under a privacy-first design: your recommendation data stays on Sauci AI only.

3. How the Recommendation Algorithms Work

Sauci AI uses a combination of:

✔ Large Language Models (LLMs)
to suggest chat styles, prompts, and themes

✔ Generative Visual Models
which adapt to the styles users interacted with previously

✔ Rule-Based Safety Filters

which prevent:

  • content resembling real humans
  • minors
  • unsafe or illegal content
  • restricted themes under our Prohibited Content Policy

✔ Preference Clustering

Our models group user selections into preference categories to provide more relevant fictional suggestions.

✔ Real-Time Adaptation

Recommendations update based on your last interactions.

4. User Controls (Required under Article 27 DSA)

You can influence or override the recommender system at any time.

Users may:

✔ Change character attributes
Age (18+ only), body type, voice, personality, style

✔ Reset recommendations
Clear interaction history → resets model suggestions.

✔ Manually choose categories
You can fully ignore recommendations and choose everything manually.

✔ Modify AI companion settings
Change preferences at any time.

✔ Disable personalized recommendations
Coming feature: “non-personalized mode” for fully neutral suggestions (DSA requirement).

5. Why You Receive Certain Suggestions

Our system may recommend content because:

  • it matches your previously selected themes
  • you interacted with similar fictional content
  • the AI model detects preference patterns
  • the content is generally popular within your category settings

This process is automated and never involves human review unless triggered by safety concerns.

6. Transparency About Limitations

Because our systems are AI-driven, recommendations may:

  • be imperfect
  • produce unexpected suggestions
  • misinterpret preferences
  • output fictional or exaggerated styles

These suggestions do not reflect real people, real-world scenarios, or actual preferences outside the platform.

7. Safety & Compliance Filters

Before any recommendation is shown, it passes through:

✔ Age-verification filters (18+ only)
✔ Likeness detection (no real persons)
✔ Minor-simulation prevention
✔ Violence, exploitation, and trafficking prevention
✔ CSAM prevention (zero tolerance)
✔ Forbidden themes filtering (see Prohibited Content Policy)

If a user attempts to request unsafe content, the system will automatically:

  • block the request
  • redirect the user
  • warn the user
  • potentially escalate to moderation

8. User Rights Under the EU DSA

EU users have the right to:

✔ Understand why recommendations appear
✔ Modify or disable recommendation logic
✔ Request human review of automated decisions
✔ Appeal a moderation action
✔ Request transparency about algorithmic processes

Requests can be sent to:

We respond within 7 business days.

9. How to Request Non-Personalized Recommendations

According to Article 27(3) DSA, users must be offered a non-personalized mode.

Sauci AI supports this by offering:

  • Neutral, random, or default character suggestions
  • Non-adaptive visual suggestions
  • No preference-based selection

Feature availability may depend on region.

10. Contact

For questions about these guidelines or DSA compliance:

Was this article helpful?

0 out of 0 found this helpful

Have more questions? Submit a request

Comments

0 comments

Article is closed for comments.