Skip to main content
Transparency

Radical Transparency — Our AI Obligations

We believe you have the right to know exactly how AI is used in your dispute, what data we hold, who reviews it, and how long we keep it.

Regulatory compliance

EU AI Act CompliantAnnex III 8(a) — ADR
GDPR Article 5Data minimisation & integrity
eIDAS-readyElectronic signatures
ADR Directive 2025/2647Certified ADR entity
ISO 27001In progress

AI Transparency Statement

Which AI model does Akordans use?

Akordans uses Claude by Anthropic — three models are deployed depending on the operation: Claude Haiku (claude-haiku-4-5-20251001) for fast lightweight operations; Claude Sonnet (claude-sonnet-4-6) for contract drafting and conversational AI; and Claude Opus (claude-opus-4-6) for deep evidence analysis and mediator briefings. We do not use GPT, Gemini, or any open-source model. Anthropic's models are subject to their published Responsible Scaling Policy. AI services for Akordans are provided by telqi.com.

What data is sent to the AI?

For evidence analysis: your uploaded documents, your written dispute statement, and the respondent's reply. Names and contact details are pseudonymised before processing. Payment data and authentication credentials are never sent to the AI.

Retention policy

AI inference logs (model, prompt version, token counts, output summaries) are retained for 3 years per EU AI Act Art. 12(1). Your dispute documents are retained for 12 months after dispute closure, then permanently deleted. You may request earlier deletion under GDPR Art. 17, subject to legal hold exceptions.

Human-in-the-Loop Guarantee

Akordans classifies each service separately under the EU AI Act:

ServiceEU AI Act RiskBasis
Free EvaluationLIMITEDAdvisory only — Art. 50 transparency
Dispute AssessmentLIMITEDAdvisory only — Art. 50 transparency
Contract ReviewLIMITEDAdvisory only — does not adjudicate between parties
Contract NegotiationLIMITEDAdvisory and drafting — does not adjudicate between parties
AI MediationHIGHAnnex III, point 8(a) — ADR applying law to a concrete dispute
EnforcementHIGHAnnex III, point 8(a) — same AI process as Mediation

For HIGH-risk services, human oversight is not optional — it is a legal requirement under Art. 14.

How we counter the HIGH-risk classification

  • Every AI output in a Mediation dispute is reviewed by the assigned certified EU mediator before it reaches either party (Art. 14).
  • The mediator's notes are attached to the audit report and visible to both parties.
  • Either party can ask, at any stage, for any AI output to be set aside and the mediator to take over directly.
  • Every AI inference (model, prompt version, token counts, output summary) is logged and retained for 3 years (Art. 12).
  • A conformity assessment and risk-management system are maintained for the AI Mediation service (Arts. 9, 16, 43).
  • You give explicit, withdrawable consent before any AI processing begins (Art. 50) and can request a plain-language explanation of any AI output.

See the full mediator review process on our Products page →

Identity Verification — eIDAS

Tier 3 disputes result in an Executory Accord — a legally binding settlement enforceable across all 27 EU Member States. For this to be valid, both parties must be positively identified.

Akordans uses eIDAS-ready Electronic Identification for Tier 3 signatories. This involves a one-time ID check (government-issued ID document + liveness check) performed by our KYC partner.

What is collected

Full legal name, date of birth, country of nationality, document number. No biometric data is stored by Akordans.

When it happens

ID verification is triggered only when both parties agree to proceed to Tier 3 and before signing the Executory Accord.

When it is deleted

ID verification records are deleted 30 days after dispute closure, unless required by applicable law or ongoing legal hold.

FAQ

Frequently asked questions

Our AI's legal knowledge

Akordans is powered by Claude, one of the world's most capable AI systems, developed by Anthropic. Claude has comprehensive knowledge of commercial law, contract principles, international regulations, and mediation practice — built from extensive training on legal texts, regulatory frameworks, case law, and professional mediation literature.

When you submit a dispute or contract to Akordans, the AI does not look up a simple FAQ. It applies its understanding of the specific legal frameworks that govern your situation — identifying applicable national laws, EU directives, international conventions, and sector-specific regulations automatically.

This is the same depth of regulatory research that a specialist law firm would conduct before advising a client. The difference is that Akordans does it instantly, and includes it in the price of every product.

Akordans' AI also understands the principles of fair mediation and context-aware negotiation — how to identify common ground, how to frame proposals that both parties can accept, and how to distinguish genuine legal disputes from misunderstandings that can be resolved through clear communication.

All AI outputs in consequential decisions — mediation resolutions, dispute assessments — are reviewed by a member of the Akordans team before being delivered to users. AI knowledge combined with human oversight is how Akordans delivers both scale and quality.

Honest about AI

Limitations and Human Oversight

We believe transparency includes being upfront about what AI does well — and where it may fall short. Here is an honest assessment.

Where AI may miss nuance

  • Jurisdiction-specific rules that post-date the model's training data (e.g. a very recent national court ruling or statutory change)
  • Novel case law in small or less-documented legal systems where training data coverage is lower
  • Party-specific equities and contextual factors that are difficult to surface in written submissions — a skilled human mediator may pick up on tone and context that text cannot convey
  • Highly technical or sector-specific regulations (e.g. pharmaceutical pricing, telecommunications spectrum) where deep domain expertise is required

What Akordans does about it

  • Dual-LLM verifier: a second AI model independently checks every legal citation for accuracy and plausibility before the output reaches you
  • Human lawyer escalation: certified EU mediators review all HIGH-risk AI Mediation outputs before delivery (EU AI Act Art. 14); you can request full human takeover at any stage
  • EU AI Act Art. 13 transparency labels: every AI-generated output is clearly labelled with the model used, its limitations, and your right to contest it
  • Full audit log: every inference is logged (model, prompt version, token counts, output summary) and retained for 3 years — so any error can be investigated, corrected, and reported

No AI system — including Akordans — is a substitute for qualified independent legal advice on complex or high-stakes matters. We say this not out of liability caution but because it is true.

Start with confidence

Every dispute is handled with full transparency, human oversight, and EU regulatory compliance.

Start a Dispute