EU AI Act

Last updated: March 2026

EU AI Act Classification

BatchCortex is designed to operate within the EU AI Act framework. Our system architecture — specifically the “Ghost Operator” human-on-the-loop model where all critical decisions require human approval — aligns with the transparency and human oversight requirements of the Act.

The precise risk classification of BatchCortex under the EU AI Act is currently under formal legal review. Regardless of classification outcome, BatchCortex is engineered to meet the more stringent High-Risk system requirements as a matter of product design:

  • Human oversight mandatory for all Critical Process Parameter decisions (Article 14)
  • Full technical documentation maintained per Annex IV standards
  • Comprehensive risk management system in place (Article 9)
  • Data governance and quality measures documented (Article 10)
  • Transparency and explainability via SHAP feature attribution (Article 13)
  • Audit trail with complete traceability (Article 12)
  • Post-market monitoring planned for production deployment (Article 72)

EU AI Act enforcement begins August 2, 2026.

BatchCortex will have a formal classification opinion from qualified Swedish legal counsel before commercial deployment.

Our approach: design for the highest standard, classify with expert guidance.

What BatchCortex Does

BatchCortex is a decision-support tool for pharmaceutical manufacturing. It monitors sensor data from production equipment, detects anomalies using machine learning (Isolation Forest), and generates AI-assisted deviation reports. All outputs are recommendations that require human review and approval before any action is taken.

Transparency Obligations

BatchCortex meets the transparency requirements of Article 50:

  • Users are clearly informed that deviation reports are AI-generated and require human review
  • Reports include confidence scores and rationale for each anomaly detection
  • The model version is recorded on every event for traceability
  • The platform labels AI outputs distinctly from human inputs in the audit trail

Human Oversight

BatchCortex is designed with human-on-the-loop architecture. No batch is released, rejected, or modified without explicit human approval. The escalation system ensures that anomalies are reviewed by QA personnel, and all batch disposition decisions require a qualified person (QP) electronic signature under 21 CFR Part 11 standards.

Roadmap: EU-Sovereign Infrastructure

BatchCortex is committed to a fully EU-sovereign technology stack. While all data is currently stored and processed within EU infrastructure, some sub-processors are incorporated in the United States with EU Standard Contractual Clauses in place. Our roadmap includes migration to fully EU-incorporated sub-processors as the European cloud ecosystem matures.

ServiceEU Data ResidencyEU-IncorporatedMigration Target
SupabaseUS parent (SCCs in place)Evaluate Hetzner / EU-native DB 2027
Mistral AIFrenchCurrent — no change needed
VercelUS parent (SCCs in place)Evaluate Hetzner/Scaleway 2027
TwilioUS parent (SCCs in place)Evaluate EU-native SMS provider 2027
ResendUS parent (SCCs in place)Evaluate Mailgun EU / Postmark EU 2027
SentryUS parent (SCCs in place)Evaluate GlitchTip EU 2027
PostHogUS parent (SCCs in place)Evaluate EU-native analytics 2027

Contact

For questions about our AI Act compliance posture or to request the full classification rationale document, contact: vilmer@batchcortex.com