What Theo refuses to do.
Architectural refusal at the workflow layer, not the chat layer. Why the platform's boundaries survive jailbreak attempts.
TheoAI is evidence-based reasoning, not ruling-making. The boundary is too important to be a marketing claim. It is engineered into the architecture — there is no surface in the platform that issues a fatwa, pronounces a ruling, or adjudicates scholarly disagreement.
The refusal lives in the architecture, not the chat
Most generative AI products are chatbots with system prompts that tell them to refuse binding output. That refusal is performative. The underlying model wants to answer. A determined user can elicit a quasi-ruling with rephrasing.
TheoAI's refusal posture is architectural. The platform has no surface that produces binding output. The reasoner surfaces evidence and dialectical context; it does not adjudicate. There is no place to jailbreak into.
Five things TheoAI will not do
TheoAI will not issue fatwas. The fatwa-issuance surface does not exist in the platform. The reasoner consults sources and surfaces classical scholarly opinions; the qualified scholar issues the ruling.
TheoAI will not adjudicate scholarly disagreement. Ikhtilāf is preserved, not flattened. The platform surfaces the disagreement; the qualified scholar engages with it.
TheoAI will not pronounce hadith authenticity beyond the corpus grading. The computational grading is published, sourced, and methodologically transparent; final authentication remains a scholarly act.
TheoAI will not advise on personal religious obligation. Questions of personal religious obligation route the user to the appropriate scholarly resources, not to a platform-issued recommendation.
TheoAI will not produce content claiming scholarly authority. The platform is decision support for scholars and students; it does not claim or perform scholarly authority on its own behalf.