AI business intelligence is legal under the EU AI Act, and as of 2 August 2026 it comes with concrete transparency obligations. The question to ask a vendor is no longer “are you compliant” but “show me”.
What the AI Act means for AI analytics tools
Under Regulation (EU) 2024/1689, an AI assistant that answers business questions is typically a limited-risk AI system. That classification carries real obligations: users must always know they are interacting with AI (Article 50), and providers must support their customers’ own transparency duties. It also has a boundary: using AI outputs for decisions like creditworthiness, insurance pricing, or hiring pushes into high-risk territory with a far heavier regime, which is why a responsible vendor prohibits those uses contractually.
How Butterflai implements it
- Classified by external counsel as a limited-risk AI system, with the analysis done by a specialized law firm, not self-assessed.
- Article 50 transparency shipped on every surface: persistent AI disclosure in chat, and “AI” markings on generated reports and dashboard widgets, on web, iOS, and Android, in all supported languages, with accessibility-grade contrast. Enforced by automated tests on the mobile apps.
- Acceptable Use Policy in our Terms of Service prohibiting use of outputs for high-risk decisions about natural persons.
- Transparency beyond labels: source-referenced answers and deterministic report pipelines, so the disclosure is backed by an architecture you can inspect. See Explainable AI.
What to ask any AI BI vendor
- Who classified your system under the AI Act, and under which risk tier?
- Where exactly does the user see that they are interacting with AI? Show every surface.
- Do your terms prevent my team from using outputs in ways that would make me a high-risk deployer?
- Where is my data processed, and is it used for training?
Our answers to all four are on this site: Trust Center and Security. Get a demo.