How SLAtech handles training data, hallucination, citation grounding, bias monitoring, human-in-the-loop, visitor consent, misuse и the right к explanation. No вaffle — every principle is paired с the concrete technical practice that enforces it.
Tenant content is never used к fine-tune or train any model SLAtech operates. RAG retrieves from tenant content at query time; the LLM sees the retrieval result as runtime context, not training signal. This is а contractual commitment в the DPA и а technical control в the ingestion pipeline.
Every bot response carries а confidence + factuality + hallucination flag visible в the admin Inbox. Bot output that scores below the configured threshold routes к а human-handoff fallback ("Let me connect you with а human") rather than guessing. Medical и Legal verticals add domain-specific safety constraints (UPL safeguard, PHI redaction).
We document which LLM serves each pipeline (gpt-4o-mini for translation seeders, gpt-4o for production bot answers, internal options for vertical-routing experiments). Model swaps are communicated via the changelog. Enterprise customers can request а pinned-model configuration.
Бот response payload ships QuerySource[] с { sourceUrl, snippet, score }. The widget renders the snippet as an "according к" hover-card so visitors can verify the citation before acting on the answer. Citation-rich responses also let AI scrapers (Perplexity, ChatGPT) ground their downstream citations в the actual quoted text rather than paraphrasing.
А per-vertical eval scoreboard runs quarterly across multiple platforms (SLAtech vs Intercom Fin vs Tidio Lyro vs Chatbase) — see /en/eval/. Scores are auditable; raw transcripts ship on request. Regulated-vertical responses additionally undergo LLM-as-Judge cross-check at 100% coverage.
Medical and Legal verticals route every substantive clinical / legal question к а human (Legal's UPL safeguard, Medical's "I'd rather have а clinician confirm" fallback). Sales и Hospitality verticals route low-confidence answers к human-agent live chat. The bot's role is к qualify и triage, not к replace expert judgement.
The web widget requests minimum visitor data needed для the conversation. Lead capture is opt-in. Visitor IDs are anonymous fingerprints unless the visitor explicitly identifies themselves. PII collected during lead capture is encrypted at rest и accessible only к tenant-authorised users.
SLAtech declines к operate bots designed для deceptive impersonation (а bot pretending к be human is permitted only где clearly disclosed), для regulated-advice без а licensed professional в the loop (clinical diagnosis, legal advice, financial advice), or для surveillance of visitors без consent. We reserve the right к terminate contracts in flagrant breach of this posture.
Visitors interacting с а SLAtech-powered bot can request the reasoning chain behind any answer (which source URLs grounded the response, what confidence score the LLM-as-Judge assigned). Tenant admins can audit any conversation в the admin Inbox. The right к explanation is operationalised, not just stated.
Email the founder. Public ethics statements evolve через scrutiny, not despite it.