Back to Washproof
SvenskaEU AI Act (2024/1689)
Ethical AI & Audit Traceability

EU AI Act & AI Transparency Statement

In accordance with Regulation (EU) 2024/1689 of the European Parliament and of the Council (EU AI Act), Washproof hereby sets out how artificial intelligence is deployed across the platform, its risk classification, data confidentiality, and Human-in-the-Loop oversight mechanisms.

Our AI Principles at a Glance

Zero Model Training

Your invoices, accounting records, and confidential business documents are never used to train, evaluate, or fine-tune public AI models.

Human-in-the-Loop (HITL)

AI proposes suggestions — humans retain decision-making authority. Any extraction with a confidence score below 0.85 is automatically flagged for manual review.

Deterministic Mathematics

Greenhouse gas emissions are never computed by generative AI. All math executes via deterministic code utilizing official factors (DEFRA, Energimyndigheten, SCB).

End-to-End Auditability

Every reported VS carbon metric traces back to the original voucher, ledger row, and supplier line.

1. Classification under the EU AI Act

Under the regulatory framework of Regulation (EU) 2024/1689, Washproof AI components are classified as Minimal to Limited Risk:

  • The Service includes no prohibited AI practices (Article 5).
  • The Service contains no high-risk AI systems (Annex III), as it performs no automated credit scoring, recruitment filtering, or judicial profiling.
  • The Service fully complies with the transparency requirements of Article 50 by explicitly notifying users whenever content or structured data has been generated with AI assistance.

2. Where and How AI is Deployed in Washproof

FeatureAI Model & RoleSafety & Verification Barrier
Invoice Lab (OCR & Extraction)Extracts fuel volumes, kWh units, and supplier identifiers from PDF vouchersConfidence scoring. Extraction confidence < 0.85 triggers manual verification queue.
NACE & Scope 3 Spend MappingSuggests industry classification codes for spend-based calculationsDeterministic fallback against official Statistics Sweden and national benchmark tables.
C1 Strategy Copilot (VS)Drafts sustainability target statements based on corporate baseline dataRequires explicit human approval and author sign-off prior to report finalization.

3. Data Confidentiality & Enterprise Infrastructure

All AI inferences are executed via Google Cloud Enterprise Vertex AI / Google Cloud Generative Language API, hosted strictly within the European Union (region europe-west1).

Under Google Cloud Enterprise terms, rigorous data isolation and confidentiality safeguards apply:

  • Customer data is not retained in persistent model memory post-inference beyond ephemeral diagnostic logs.
  • Customer inputs are never shared with third parties nor incorporated into foundation model training loops.
  • All transmissions are secured over TLS 1.3 with strict API authentication.

4. Point of Contact for AI Governance and Audits

Do you or your sustainability auditors have technical inquiries regarding our calculation methodologies, algorithm transparency, or validation criteria? Contact our compliance department at compliance@washproof.eu.

© 2026 Washproof (Softstuff AB). All rights reserved.
Privacy PolicyTerms of ServiceGDPR DPA