AI and Automated Systems Notice

Status: Counsel-approved production version.

Operator / contracting entity: UAB Skillaxis, trading as InquOS

Company code: 307395574

Registered office: Architektų g. 56-101, LT-04111 Vilnius, Lithuania

Legal and support contact: support@sufoniq.com

Privacy contact: privacy@sufoniq.com

This Notice explains how InquOS may use automated and AI-supported systems. It must be published as a standalone Legal Center page and supplemented by concise in-context notices where a feature is used. It must be updated from the actual AI feature register; a generic statement that “we use AI” is inadequate.

G1. What these systems may do

• extract and structure information from resumes, identity documents, certificates and other uploads;

• identify possible duplicates, inconsistencies, missing evidence, stale information or fraud indicators;

• map occupations, skills, qualifications and experience to structured taxonomies;

• suggest mobility pathways, evidence needs, next steps or questions for review;

• rank or recommend jobs, candidates, education options or service providers based on disclosed parameters;

• generate summaries, explanations, draft communications and operator-support material;

• support safety, security, quality assurance and platform-integrity review.

G2. What these systems do not decide for InquOS

• They do not issue a visa, permit, licence, qualification, admission or government decision.

• They do not guarantee a person is legally eligible, employable, admissible, suitable or trustworthy.

• They do not replace a qualified professional where legal or regulated advice is required.

• They must not be used by a business user as the sole basis for a significant employment, admission, professional or service decision where law or fair process requires independent review.

G3. In-context transparency

A feature that directly interacts with a user through AI or displays materially AI-generated content should identify that fact clearly unless it is obvious from the context. Generated or transformed content must be labelled at the point of use where the label is needed to avoid confusion. The user should be able to distinguish:

• facts supplied by the user or an authoritative source;

• structured extraction or classification produced by a system;

• a recommendation or inference;

• AI-generated draft wording;

• an operator or professional decision;

• an unresolved conflict, uncertainty or review requirement.

G4. Correction and human review

Users must be able to correct material factual information. Where an automated result materially affects platform access or a significant workflow, the product must provide a route to request human review. The reviewer must be able to assess the relevant facts and must not merely repeat the system output. Security and fraud controls may limit disclosure of detection methods, but not eliminate a meaningful correction route where one is legally required.

G5. Data use and model providers

The Privacy Notice must identify the categories of information sent to model or AI service providers, the purpose, legal basis, retention, access rules, transfer mechanism and whether the provider uses the information for its own purposes. InquOS must contractually and technically minimize provider retention and training use where the service design requires confidentiality.

Product commitment to confirm before publication

Recommended wording only if technically and contractually true: “Unless separately and expressly agreed, InquOS does not use private uploaded documents, private case files or private user communications to train general-purpose AI models.” Do not publish this sentence until every relevant vendor configuration and contract supports it.

G6. Responsible operation

• Maintain an internal register of every automated or AI-supported feature, owner, purpose, users, data, model/vendor, output, decision impact, risks, controls, evaluation and legal classification.

• Test extraction accuracy, false positives, subgroup performance, accessibility and meaningful edge cases before material deployment and after significant changes.

• Use confidence, provenance, effective-date and uncertainty controls appropriate to mobility and evidence reasoning.

• Prevent silent guessing where official information is incomplete. Unsupported or ambiguous cases must be escalated, limited or clearly marked.

• Keep appropriate logs and version records so a material output can be reconstructed without retaining more personal data than necessary.

• Train employees and operators who use AI-supported outputs and assign clear responsibility for final human decisions.

• Provide incident escalation for materially wrong, discriminatory, unsafe or privacy-invasive outputs.

G7. Feature-classification register

Required field

Example of what InquOS must record

Feature and owner

Resume parser — UAB Skillaxis — InquOS Product & Engineering.

Purpose and affected users

Structures candidate-provided facts for review; affects job seekers and operators.

Model / provider

gpt-4o-mini, provided by OpenAI.

Input and output

Uploaded document; extracted facts with provenance and confidence.

Decision impact

Informational / recommendation / access-affecting / significant-decision support.

Human oversight

Correction by user; operator review before route activation.

Legal assessment

GDPR profiling/Article 22; AI Act role; consumer/employment/discrimination implications.

Evaluation

Accuracy by document type/language, false conflicts, subgroup and edge-case tests.

Retention and audit

For the Chat Completions API, prompts, responses and related information may be retained in abuse-monitoring logs for up to 30 days by default, unless stricter approved retention controls apply or longer retention is legally required. Structured extraction results retained by InquOS are kept with the associated document, account or workflow only for as long as reasonably necessary under the InquOS Privacy Notice; content hash, model version and review result are retained as applicable.

User notice

In-context label and link to this Notice.

G8. Current implementation timeline

As at 27 July 2026

AI Act transparency obligations relevant to user interaction and AI-generated content apply from 2 August 2026. Following Regulation (EU) 2026/1744, the principal Annex III high-risk obligations are scheduled to apply from 2 December 2027. Feature classification remains essential because an employment-related system may be high-risk depending on its intended purpose and use. GDPR, discrimination, labour, consumer and safety duties apply independently of that timetable.