1.1. This Notice explains how artificial-intelligence and automated systems support the Employza Service, what Users can expect, the main inputs and outputs, important limitations, human-oversight requirements and available controls. 1.2. Employza is a modern global AI work platform that uses artificial intelligence, including agentic systems, to support people and organisations in work, career development, recruitment and workforce operations. The Service is operated by Krzysztof Rąpała under the Employza brand. Contact: support@employza.pl. 1.3. This Notice supplements the Terms of Service and Privacy Policy. The Privacy Policy governs legal bases, recipients, transfers, retention and data-subject rights. The current product interface and authenticated capabilities determine which AI functions are actually available.

2.1. Employza identifies direct AI interactions through product names, interface labels and contextual notices, including Vision, Vision Advisor, Vision Intelligence, QuickShift, Interview AI, Voice Search AI, AI Profile Sync and Hiring Copilot. 2.2. Output generated or assisted by AI may include messages, summaries, recommendations, rankings, extracted data, draft documents, interview practice, public-web research or proposed actions. Where applicable law requires an additional label for generated or manipulated content, Employza applies that disclosure in the relevant output or publication workflow. 2.3. An AI message is not proof that an account action occurred. Only a verified action result, application record, billing record or other trusted Service confirmation establishes completion.

Employza may use AI or automated processing for: (a) job search, structured query interpretation, matching and recommendation; (b) candidate suggestions for an Employer's internal offer; (c) internal applications, Smart Apply planning and application-status support; (d) Hiring Copilot analysis for eligible internal applications; (e) QuickShift candidate discovery, comparison, staffing plans and controlled contact reveal; (f) AI Profile Sync, Document Matching AI, CV extraction, scanned-document reading and professional document generation; (g) Vision Chat, Vision Advisor, Vision Intelligence and administrative assistance; (h) public-web research with source links; (i) Voice Search AI, realtime voice interactions and Interview AI practice; (j) optional Personal Memory for conversational continuity; (k) content quality, deduplication, translation, classification and aggregated trend analysis; and (l) security, fraud, abuse and anomaly detection. This catalogue describes product categories, not internal models, prompts, weights, thresholds, infrastructure or protected implementation know-how.

4.1. Purpose: reduce the time needed to identify potentially relevant jobs or candidates while preserving User choice and human judgment. 4.2. Typical inputs: job title and description; required and preferred experience; skills and tools; work mode; location; language; employment preferences; candidate profile and CV-derived professional information; availability; and current authorised workflow context. 4.3. Typical outputs: ordered job results, candidate suggestions, a match indication, reasons for fit, comparison points or near-match information. 4.4. Limitations: matching is probabilistic. It may miss a suitable opportunity or surface an unsuitable one because source data is incomplete, language is ambiguous, requirements change or professional experience cannot be fully represented. It is not a verification of identity, qualifications, availability or legal eligibility. 4.5. User control: Candidates may edit profile information, change filters, provide another document, disregard a recommendation or choose an identifiable current internal offer that was not recommended. Employers must independently assess candidates.

5.1. Vision Advisor may resolve candidate-selected or matched offers and prepare a bounded internal application plan. Smart Apply lets an eligible Candidate select several internal Employza offers while browsing and continue to the same review and confirmation process. 5.2. Before execution, the Candidate sees or receives the exact internal selection and must explicitly confirm it. The system records the result per offer. Already-submitted, inactive, blocked or failed items do not cause a false success statement. 5.3. External offers are separated before confirmation and identified for manual completion. Vision does not open, prefill, submit or retry an external recruitment-system form. Employza does not infer completion from a click or from a Candidate's independent activity outside the Service. 5.4. AI may help prepare truthful application material, but the Candidate remains responsible for the facts and final selection.

6.1. Hiring Copilot provides an Employer with advisory support for an eligible internal Employza application. It may generate five areas: a concise candidate overview, role-specific fit, strengths, matters requiring verification and suggested interview questions. 6.2. Typical inputs are the internal job offer, the Candidate's submitted CV or application material and structured professional signals lawfully available in that application workflow. 6.3. Hiring Copilot does not accept, reject, contact or make an employment decision. The Employer must verify the evidence, conduct a fair assessment and preserve human oversight. 6.4. Hiring Copilot does not run for external-link applications because the receiving recruitment process is outside Employza.

7.1. QuickShift helps authorised Employers describe an urgent staffing need, search eligible participating Candidates, compare profiles, build shortlists and prepare staffing actions. 7.2. Typical inputs include the Employer's stated need and professional, availability, location, language and preference information of Candidates who currently satisfy the QuickShift participation and visibility conditions. 7.3. Candidate contact details remain hidden until an authorised reveal is expressly requested, confirmed and permitted by the current plan. AI output cannot override consent, visibility or allowance rules. 7.4. QuickShift does not verify that a Candidate will accept an engagement and does not conclude an employment or services contract for the parties.

8.1. AI Profile Sync and document features may extract, classify and organise information from readable or scanned CVs and other supported files; derive job-search criteria; propose profile updates; or generate a CV, cover letter, job offer, HR draft or other professional material. 8.2. Typical inputs are the User's file, profile, instruction and authorised job or organisation context. Typical outputs are structured fields, search criteria, summaries, suggested edits and downloadable or editable documents. 8.3. Generated content is probabilistic and may contain errors, omissions or unsuitable wording. Employza instructs the system not to invent candidate facts, but the User must verify identity, experience, education, competencies, dates, legal terms and every statement before use. 8.4. A draft contract, policy or HR document is not legal advice and does not become binding merely because AI generated it. Appropriate professional and organisational review remains necessary.

9.1. Vision Chat is a conversational service without private account access or action-taking authority. Eligible Vision Advisor, Vision Intelligence and administrative Agents may receive server-authorised tools appropriate to the authenticated role and plan. 9.2. The Agent may analyse context, prepare a plan or execute a supported action only within current permissions. Consequential actions require the confirmation or authorisation shown by the Service. The system rejects tool requests outside the authorised inventory even if requested in a prompt. 9.3. Eligible Employer workflows may use organisation-scoped assignments, approvals, notes, interview events, recorded outcomes and shared operating rules. These records belong to the organisation workspace, are distinct from an individual's optional Personal Memory and remain subject to organisation access controls and the Employer's recruitment and data-protection duties. 9.4. Public-web research may search current public sources and return source-backed findings. A citation identifies the source used; it does not guarantee that the source is accurate, complete or unchanged. 9.5. Background tasks may preserve progress and report status where supported. They do not broaden the User's instruction and are not the same as an unreleased or unavailable product capability.

10.1. Voice Search AI converts a deliberately recorded spoken request into text and structured job-search criteria. It is not speaker identification and does not create a persistent voiceprint. 10.2. Realtime voice enables a live AI conversation in eligible Vision or QuickShift surfaces. The same role, plan, data-access, confirmation and tool boundaries apply as in text. 10.3. Interview AI is a Candidate-only practice simulation. It may adapt questions and feedback to the conversation and selected professional context. It is not a real employer interview, a psychological assessment or an employment decision. 10.4. Speech recognition and generated responses may mishear, misunderstand or oversimplify. Users should correct material errors and should not rely on the system for emergency or regulated professional advice. 10.5. Realtime usage limits are calculated using started minutes. Each session uses at least one minute, and additional partial minutes are rounded up.

11.1. Personal Memory is available only on eligible signed-in surfaces and is off until the User explicitly enables it in Account settings. 11.2. It stores selected non-sensitive preferences and continuity facts provided by the User or expressly requested to be remembered. Only the smallest relevant set is loaded for a conversation. 11.3. Users can inspect, add, correct and delete items, pause Memory or reset it. Pausing stops use and creation while preserving items for control; reset deletes them and turns Memory off. Each item expires no later than 365 days after creation or update. 11.4. Memory may be incomplete or outdated. It is not verified evidence of identity, qualifications, employment history or consent to an action. It never grants a role, plan, tool, contact access, confirmation or other authority.

12.1. Employza's recruitment-related AI is designed as information, preparation and decision support. Employza does not permit AI output alone to accept or reject a Candidate, terminate employment, set employment terms or make another legal or similarly significant employment decision. 12.2. Employers are responsible for meaningful human review, lawful criteria, evidence verification, accessibility, equal treatment and the final decision. Candidates remain responsible for their application choices and submitted facts. 12.3. Certain intended uses of AI in recruitment or worker management may fall within high-risk categories under the EU AI Act or other laws. Employza assesses intended use and deployment. Where a legally applicable classification triggers additional requirements, the relevant system must satisfy those requirements before that use is released or materially expanded. 12.4. A narrow procedural or preparatory function is not automatically harmless. Employza evaluates actual purpose, influence, profiling and foreseeable effect rather than relying only on a feature label.

13.1. Employza aims to use relevant professional criteria and not protected characteristics for matching or recruitment support. Users must not request or use AI to discriminate on the basis of race, ethnic origin, nationality, sex, pregnancy, religion, disability, age, sexual orientation, gender identity, trade-union activity or another protected status. 13.2. Bias can arise from incomplete profiles, job wording, historical patterns, language coverage, source quality or model limitations. We use proportionate testing, monitoring and review, but no system is error-free or universally accurate across every profession, country and language. 13.3. Users can correct source information, change criteria, disregard output, report a concern and request appropriate human review. Employza may restrict a feature or output where safety, fairness or legal compliance cannot be maintained. 13.4. Accessibility needs and non-traditional career paths should be considered by human reviewers and not treated as negative evidence merely because they are less represented in structured data.

14.1. Employza minimises AI inputs to the information reasonably needed for the requested feature. Account role, organisation scope, plan, consent, confirmation and access rights are determined by trusted server-side state, not by a User prompt or Personal Memory. 14.2. Employza may use contracted providers of AI processing, hosting and infrastructure. They receive only data needed for the instructed service and are subject to contractual, confidentiality, security and data-protection requirements appropriate to their role. 14.3. Employza does not use private User Content to train a general-purpose AI model unless a separate, clear notice and valid opt-in expressly authorise that use. Service-quality evaluation may use minimised, access-controlled or anonymised information where lawful. 14.4. International transfers and retention are governed by the Privacy Policy. Protected prompts, routing, model configuration, thresholds, algorithms, infrastructure and security controls are confidential and are not disclosed by this Notice.

15.1. Automated systems may support detection of spam, fraud, credential abuse, scraping, malicious files, anomalous traffic and unsafe requests. These controls protect the Service and Users; they do not authorise covert employment decisions. 15.2. Users must not manipulate an AI system to obtain another person's private data, evade confirmation, reveal restricted candidate contacts, generate unlawful content, test production security or reconstruct protected internal mechanisms. 15.3. A safety or quality concern may be reported to support@employza.pl. Employza may preserve relevant evidence, restrict an output or feature, investigate and notify competent authorities where required. 15.4. A security signal may be wrong. Where a restriction materially affects a User and law requires it, an explanation and human review are available, subject to lawful limits protecting security and other persons.

16.1. Privacy rights, including access, correction, deletion, objection and relevant automated-processing information, are described in the Privacy Policy. Product controls allow Users to edit source data, manage conversations and Memory, control consent and decline recommendations. 16.2. AI output should be challenged when it appears inaccurate, discriminatory, unsafe or unsupported. A report should identify the feature, approximate time, affected output and concern without sending unnecessary sensitive data. 16.3. We update this Notice when an AI function, intended use, oversight model or applicable transparency duty materially changes. The current date appears at the top. Material disclosures are provided before or when the relevant change becomes available. 16.4. Translations are provided for accessibility. Mandatory local rights prevail. Questions and reports may be sent to support@employza.pl.