One format, every sector where AI decides

Wherever an AI system makes or shapes a consequential decision, someone will eventually ask why. The Causal Seal answers — in finance, health, justice, mobility and beyond.

How to read these cards. The six causal classes (identity · engine · field · shaping · context + timestamp) are universal. Each sector then defines a domain profile: which specific parameters a deployment in that field should seal. The profiles below are illustrative proposals, offered to start the conversation — defining them rigorously is open work for each sector's community, not a claim that these exact fields are already standardized.

💳 Finance & market algorithms

Prove the compliance of an investment or credit decision (MiFID II, AI Act) without exposing the secret strategy behind it.
The seal fixes the risk constraints engaged (shaping.*) and the market regime (field.regime) at the instant of the order — auditable evidence without revealing the model or the strategy.

🩺 Health & medical diagnosis

Give health authorities rigorous traceability for AI-assisted diagnoses and therapeutic recommendations.
The seal records which clinical expert profile was engaged (identity.expert) and exactly how much patient context was visible to the model (context.*) at the millisecond of the recommendation.

⚖️ Justice & legal assistance

Guarantee that a legal-assistance AI reasons only from the supplied statutes, with creativity constrained.
A domain profile can seal a digest of the legal sources actually consulted (context.*) and the constraint state that suppressed free invention (shaping.*) — making undetectable hallucination auditable.

🚗 Transport & autonomous vehicles

Provide an indisputable "semantic black box" to reconstruct an avoidance decision after an incident.
An automotive profile can seal a hash of the perceived sensor frame (context.*) and the weight given to pedestrian safety (shaping.*) at the precise moment of the manoeuvre.

⚡ Energy & critical infrastructure

Trace, instantly, an AI's load-shedding or grid-balancing decisions.
A profile can seal the sensor telemetry the model acted on (context.*) and the physical safety margins imposed on it (shaping.*) at the millisecond of the decision.

📱 Content moderation & platforms

Justify a removal or ban (DSA) by objective criteria, free of hidden algorithmic bias.
The user can receive a seal proving the tolerance threshold applied and how context (e.g. irony) was weighed (shaping.*) — a neutral basis for appeal and audit.

🔐 Web3, crypto & autonomous agents

Record an AI agent's decision history on-chain, cheaply and tamper-proof.
Because the seal's fingerprint is tiny, it anchors at near-zero cost inside on-chain transactions or Merkle trees — decentralized proof of decision integrity.

🛡️ Defense & safety-critical autonomy

Guarantee human accountability and traceability of the criteria behind AI-assisted actions in high-assurance systems.
The seal fixes the activated expert (identity.expert), its certainty (identity.confidence) and the ethical safety guard (shaping.guard) before any action — so responsibility stays assignable to a human, and no autonomous action escapes audit.

→ How a seal works · → Our neutrality charter · Define a domain profile with us

Causal Seal — an open standard for decision provenance. Sector profiles shown are illustrative and open for community definition. Informational only; not legal advice. DOI 10.5281/zenodo.21431267.