JBKR Web uses artificial intelligence in two places: as an internal development tool and as part of the solutions it delivers. This page states, in plain language, where AI comes in, what it never does with your data and who is accountable for decisions. Five principles, dated and signed, reviewed every year or whenever the law changes.
Updated on
Client data (source code, databases, documents, user conversations) is not used to train AI models, ours or anyone else's. Integrations use commercial APIs in modes that do not retain data for training, or private models run on the client's infrastructure when data cannot leave the premises.
No AI output reaches production, your codebase or a business decision without review by a qualified person. In delivered solutions, critical flows (payments, personal data, decisions that affect customers) require human approval (human-in-the-loop). The person accountable for the work is JBKR's founder, not the model.
Every AI solution ships with an evaluation set (evals) run on every prompt, model or knowledge-base change, input and output guardrails, and quality, cost and latency telemetry in production. If it cannot be measured it cannot be promised, and JBKR does not promise it.
Projects follow Brazil's LGPD (Law 13,709/2018) in the processor role, track Brazil's AI bill (PL 2338/2023) and use the NIST AI Risk Management Framework and the EU AI Act as references to classify risk. High-risk use cases (automated decisions on credit, health, employment) are only accepted with mandatory human oversight and a documented impact assessment.
Internally, coding assistants speed up repetitive tasks, reading legacy systems and writing documentation. What goes into those assistants is controlled: no credentials, no end-user personal data and no unmasked client databases. The productivity gain is passed on to the client in lead time and price, not hidden.
Jean C Becker, founder of JBKR Web (WEB ROOTZEN TECNOLOGIA LTDA) — Curitiba, Brazil, 5 September 2026.
