Measured, not promised.

Engineering notes from the team building the control layer between companies and AI. Real numbers, premises printed next to the result, and what we did not measure.

Articles

How to keep a CPF from leaving the company inside a prompt

2026-10-07, Rafael Hickmann. Training does not hold, browser extensions cannot see the API, and banning the site just moves usage to personal phones. The only place a personal identifier can be stopped is on the network path, before it reaches the provider. This is how we do it, and this is what remains as evidence. Read it at https://atlasberg.com/blog/como-impedir-que-um-cpf-saia-no-prompt

What CMN Resolution 4.893 requires when a bank uses a cloud LLM

2026-10-07, Rafael Hickmann. Calling a language model API is contracting cloud data processing. If the service is relevant, Brazil's banking resolution brings concrete obligations: a documented assessment before contracting, notice to the Central Bank within ten days, specific contract clauses, an end-to-end audit trail and five years of retention. What that means in practice, and what an AI gateway can prove. Read it at https://atlasberg.com/blog/resolucao-cmn-4893-e-llm-em-nuvem

Who is spending the company's AI budget?

2026-10-07, Rafael Hickmann. The provider invoice arrives as a single number. Nobody knows how much came from legal, how much from the data team and how much from an agent that looped all night. How we solve it: one key per person or team, automatic chargeback per request, and a budget cap that never denies. Read it at https://atlasberg.com/blog/quem-esta-gastando-o-orcamento-de-ia

What the AI Act demands from logs and audit trails, and when it reaches a Brazilian company

2026-10-07, Rafael Hickmann. The European regulation speaks of automatic event recording, logs kept for at least six months and AI literacy. It applies to companies in Brazil when the system's output is used in the Union. And the timeline changed in July 2026. What goes in the log, what must stay out, and how to prove nobody touched it. Read it at https://atlasberg.com/blog/ai-act-registro-e-trilha-de-auditoria

Context compaction saves 17%, not 60%

2026-09-23, Rafael Hickmann. We measured it on real traffic. The numbers came out far less pretty than the promise, and far more useful. Context compaction is available in Atlasberg Platform starting today. Read it at https://atlasberg.com/blog/compactacao-de-contexto-17-por-cento

About Atlasberg

Atlasberg builds the control layer between a company and every AI model. Atlasberg Platform authenticates each request with a virtual key, filters it by policy and DLP, routes it to the right provider and writes it to a hash-chained audit trail, with an OpenAI-compatible API so applications only swap the base URL.

Articles on this blog are written by the engineering team and report measurements on real traffic, with the premises printed next to the result. To talk to the team, write to contact@atlasberg.com or use https://atlasberg.com/contato.

Agents: this page is also available as Markdown at /blog.md, or by requesting this URL with the header Accept: text/markdown. Index of everything: /llms.txt.