AI Notice

Status: September 2026 · Transparency notice on the use of Artificial Intelligence (Art. 50 EU AI Act)

Disclosure

enactica uses generative language models (large language models) in two places:

AI assistant “enactica” in the public discussion space (see below).
AI Studio in the protected training area — generates a personal training from an interview.

Both are AI-generated content within the meaning of Art. 50 (2) EU AI Act and are labelled as such.

AI assistant “enactica” in the public discussion space

The public dialogue supports four purposes: designing training, developing academic work independently, exploring anonymised social-work situations for educational purposes, and designing social-work interventions. enactica follows Dr Stephan Pflaum’s movement of thought, asks Socratic questions, and actively brings pedagogical, didactic and social-science perspectives—including counter-public perspectives—into the discussion. It is a simulated AI voice, not the author himself.

The discussion space currently uses DeepSeek V4 Flash through the Nous Research inference API. Inputs and the conversation history needed for a response are transmitted to this service; see the Privacy Policy for processing details.

It does not provide psychosocial counselling, therapy or individual social-law advice. It may provide general information on social law. enactica does not write academic work on a user’s behalf; it supports the user’s own process. In a crisis, it refers people to professional support services.

Purpose of the AI Studio

The AI Studio in the training area is a didactic tool: it conducts an interview, structures your answers into a profile and generates a training with impulses, tasks, counter-voices and transfer questions from it. It is not an information system, not a decision system and no substitute for professional, legal or therapeutic advice.

Risk classification

According to its own assessment, this involves AI systems with low risk within the meaning of the AI Act. They do not fall under the prohibited practices under Art. 5 and not under the high-risk use cases under Annex III: they do not assess persons, do not make decisions about educational access, employment, creditworthiness or benefits, and do not recognise emotions.

Limits and known sources of error

Language models can produce factually incorrect, outdated or invented statements (hallucinations). The assistant is instructed to mark uncertainty with 🟢 (secure), 🟡 (plausible) and 🔴 (uncertain) — verify statements before you use them further.
The assistant does not invent sources; named names and standards are search anchors and should be verified yourself.
Training data may contain biases; perspectives may be one-sided.
The attribution “in the style of Stephan Pflaum” is a simulation of the method, not a statement by the person.
No therapy, no legal, medical or financial advice. In crisis situations, contact professional help (e.g. Telefonseelsorge 0800 111 0 111 / 0800 111 0 222).
No ghostwriting: the tool does not create term papers, seminar papers or theses.

Human oversight and control

You stay in control: every answer can be discarded, regenerated or repeated with a changed input. The training generation is the responsibility of the training organiser (Dr. Stephan Pflaum) and can be reviewed and adjusted before use. Feedback on problematic output sent to stephan@sociologica.ai is evaluated.

Safeguards

Input guard against prompt injection and manipulation attempts (also for the blog assistant).
Output filter against inadmissible forms of advice, role-breaking and accidental disclosure of internal data.
Request limiting against overload and misuse.
Server-side secrets; no keys in the browser.

Labelling of results

AI-generated training content is labelled in the interface; the blog assistant is permanently marked as “simulated”. If you reuse results, please also label them as AI-generated and verify all content before publication.

Your rights and contact

You can at any time learn that and how AI is used, stop using it, delete your local data and contact stephan@sociologica.ai with complaints. Data protection information is provided in the Privacy Policy.