Documents are not proof
A policy claim, a defined procedure, technical evidence, and an operational record do not provide the same strength of support.
Policies and supporting records are scattered across governance tools and repositories. Reviewers still have to find the relevant statement, judge its strength, connect it to a requirement, resolve inconsistencies, and explain the conclusion later. That manual reasoning is slow, difficult to reproduce, and easy to lose inside a conventional audit trail.
A policy claim, a defined procedure, technical evidence, and an operational record do not provide the same strength of support.
Different sources can describe the same system, actor or control differently—and the inconsistency matters.
Keep the governance and document workflows your team already uses. Certaelis adds a reasoning layer between source material and regulatory conclusions, with a visible basis for every step.
Explore the workflowCertaelis connects each conclusion to the underlying facts, exact citations, applicable requirements, conflicts, review decisions, and saved assessment history.
Extract individual candidate facts from existing PDFs and keep the supporting quote, document, and page attached.
Trace every claim back to its exact source.Use source-linked facts to prefill applicable EU AI Act requirements while keeping missing support visible for review.
See what the evidence supports—and what it does not.Surface contradictory candidate values and hold them out of the assessment until a reviewer records a resolution.
Turn ambiguity into a documented decision.Keep saved findings tied to their regulatory release, source evidence, review status, and documented decisions.
Reconstruct the basis of a historical assessment.The reasoning chain stays explicit from source material to atomic facts, applicable requirements, human approval, and the saved result.
Start an audit under a selected regulatory release, then add relevant policies, procedures, contracts, technical files, and governance records as PDFs.
Input: the documentation you already maintainThe extraction workflow identifies individual candidate facts and returns exact source quotes, document references, page locations, and warnings.
Output: atomic facts with provenanceSource-linked facts populate applicable article templates and assessment fields, while missing or insufficient information remains visible.
Output: requirement-level evidenceReviewers accept, correct, or reject extracted values, compare conflicting sources, and record explicit decisions.
Output: a human-reviewed fact setEvaluate applicable requirements and separate supported findings from partial coverage, missing facts, conflicts, and open review work.
Output: auditor-ready evidence and gapsEach step preserves the link between source documents, atomic facts, applicable requirements, contradictions, and human decisions. A reviewer can interrogate a finding all the way back to the exact citation behind it.
Capture the facts needed to classify an AI system through a guided review. Certaelis organises the relevant information, supports the reviewer’s decision, and connects the resulting classification to the applicable EU AI Act obligations.
Provide PDF documents and let the extraction workflow identify candidate facts with source quotes and page references.
Accept, correct or reject extracted values before they are relied on in an assessment.
Compare contradictory candidate values, inspect their evidence and save a documented resolution.
Keep saved findings tied to their regulatory release, source evidence, review status, and documented decisions.
Assess applicable requirements individually and keep missing facts, conflicts and non-compliant findings visible.
Download a saved assessment result with its review sections in a portable report.
The platform prepares evidence-backed candidate values and assessment fields. It does not turn an AI-generated output into a final legal or compliance decision.
Every answer remains under human control.
Move beyond a single score. Distinguish supported findings from partial coverage, missing facts, contradictory sources, and conclusions still waiting for human review.
The result helps reviewers prioritise follow-up. It is not a compliance guarantee or an autonomous legal conclusion.
Illustrative workspace data. Statuses use text and icons in addition to colour.
When documents disagree, Certaelis keeps each active candidate value and its source visible until a reviewer records an explicit resolution.
Interface content is illustrative and contains no customer data.
Can oversight personnel decide not to use the system or to disregard, override, or reverse its outputs?
Select the value supported by the audit evidence.
No candidate is used by compliance while this conflict is unresolved.
Support internal governance teams and external reviewers with a traceable path from supplied documentation to findings—without asking them to replace established GRC and document workflows.
Give legal, compliance, risk, and system owners a shared way to prepare internal or external assessments without losing the source behind a conclusion.
Use a consistent, structured workflow to review supplied documents, focus professional judgment where it matters, and deliver clearer findings.
Each audit is tied to the immutable regulatory release selected when classification begins. That basis flows through extraction, requirement mapping, human review, and the saved assessment result.
Inherited from classification and retained with the saved assessment result.
eu-ai-act@1.0.0templates/1.0.0Recorded · 7b3f…a91cPlans and pricing are being shaped around the scope, frequency, and collaboration needs of real assessment teams.
We’ll use that conversation to understand the right access model for individual reviews, recurring team work, or wider governance programmes.
No purchase is required to discuss your workflow.
Certaelis supports structured review. It does not turn an automated output into professional advice.
Discuss your workflowCertaelis is a compliance evidence workspace for EU AI Act reviews. It turns supplied PDFs into atomic, source-linked facts, maps them to applicable requirements, surfaces missing or contradictory evidence, records human decisions, and exports saved results for review.
It is designed for people involved in assessing AI systems: legal and compliance professionals, AI governance or risk teams, external advisors, and the product or control owners who provide the underlying facts and evidence.
The current extraction workflow accepts relevant PDF documents supplied as audit evidence. These may include policies, procedures, contracts, technical documentation, governance records, and risk documentation. Direct document-repository integrations are not yet available.
It identifies candidate facts in supplied documents, connects those facts to source evidence, and prefills selected assessment templates. Reviewers can focus on checking evidence, resolving conflicts, and making decisions instead of repeatedly searching and copying.
No. Extracted values can be accepted, corrected, or rejected, and conflicting evidence can be escalated for resolution. The product is designed to support a qualified reviewer, not remove that reviewer from the process.
When candidate values disagree, the platform presents the active values with their evidence. A reviewer can select a supported candidate or enter a different value, document the rationale, save the resolution, and revisit its history.
An audit keeps the regulatory release selected during classification, along with a record of the regulatory content used. A later release therefore does not reinterpret a historical result. Versioning provides traceability; it does not guarantee compliance.
Yes. A saved assessment result can be downloaded as a PDF, including sections for items needing review, conflicts, missing facts, and non-compliant findings.
Plans and pricing are being finalised around the needs of early users. Book a demo to discuss the scope and frequency of your assessment work.
Submit the demo form below with a short description of your workflow. If requests are temporarily closed, the page will say so before collecting any details.
Tell us about the evidence-heavy reviews you run and the documentation your team already maintains.
Professional review remains essential. AI-generated and system-generated outputs should be reviewed by a qualified person. Certaelis does not automatically provide legal, accounting, compliance, regulatory or other professional advice.