AI transparency

How we use AI

The core of CounterBrain is the "contradictor": a set of AI agents that analyse a decision or hypothesis by actively searching for implicit assumptions, counter-intuitive scenarios, and falsification tests. The goal is not to confirm the user's hypothesis but to subject it to rigorous stress-testing.

The AI models operate behind a multi-provider gateway: the user does not interact directly with a single model but with an architecture that can orchestrate multiple LLMs (Claude by Anthropic as default) depending on the type of analysis. This approach allows us to remain model-agnostic and to operate in environments with data sovereignty requirements (on-premise / air-gapped).

No automated decisions

CounterBrain does not make automated decisions within the meaning of Art. 22 GDPR that produce legal or similarly significant effects on users. The system is a critical-thinking support tool: it generates analyses, counter-arguments, and falsification questions, but the final decision always rests with the human.

We do not build user profiles nor use user behaviour to influence the contradictor's output in an opaque way.

Limitations

• Contradictor outputs are probabilistic: the model may make errors, hallucinate facts, or produce partial analyses. We always recommend verifying cited sources and subjecting analyses to independent critical review.
• The system does not have real-time access to all available information: output quality depends on the quality of the input provided by the user and on the knowledge embedded in the model at its training cutoff date.
• Human oversight is a non-negotiable requirement: CounterBrain is designed to amplify human judgement, not to replace it.

Sources and provenance

When the contradictor cites external sources (web search, retrieved documents) it indicates them explicitly in the output with provenance and retrieval date. Coopera S.a.s.'s proprietary dataset — built through real contradiction cycles — is exclusive and non-replicable: it constitutes the platform's primary long-term competitive advantage.

For a detailed description of the contradiction methodology, the three-agent architecture, and the Δ-CSI (Cognitive Sovereignty Index) calculation, see the How it works page.