Skip to content
WorkSettle

Legal

AI Use & Transparency Policy

How WorkSettle uses artificial intelligence, which data it processes, what is never used for training, and the controls you have.

1. Purpose and Our Commitments

WorkSettle Inc. ("WorkSettle", "we", "us") uses artificial intelligence ("AI") to make Canadian immigration and job search easier to understand and navigate. Immigration decisions affect people’s families, careers and futures, so we hold our AI to a high standard of honesty about what it does, what data it uses, and where it can go wrong.

This Policy explains, in plain language, every place we use AI on the WorkSettle platform (the "Platform"), which models we use, what data the AI collects and processes, what data is and is not used to train or improve AI, and the controls you have. It applies to Users, Businesses and Professionals (RCICs, lawyers and other service providers) and forms part of our Terms & Conditions and Privacy Policy.

Our commitments at a glance.

  • We tell you when you are dealing with AI. AI-generated content is labelled, and the AI Agent never pretends to be a human, an RCIC, a lawyer or a government officer.
  • AI informs; people decide. Our AI provides general information and tools. It does not make legal, immigration or hiring decisions about you, and it is not legal or immigration advice.
  • Your documents and conversations are not used to train AI by default. We use identifiable data for AI training only with your separate, express opt-in consent, which you can withdraw at any time.
  • Our AI providers may not train on your data. We contract with model providers who are prohibited from using WorkSettle data to train their own models.
  • Professional and Business data is never used for training. Client files that RCICs, lawyers and Businesses handle on WorkSettle stay confidential to them.
  • We ground answers in official sources and show you where information comes from, so you can check it.
  • We do not sell personal information, including AI conversations, to anyone.
  • You stay in control: you can view, export and delete your AI history and change your AI settings at any time.

2. Where We Use AI on the Platform

The table lists every AI feature on the Platform. If we add a new AI feature, we will add it here before it launches.

AI features on the WorkSettle platform
FeatureWhat the AI doesWhat it does not doHuman role
AI Immigration Agent (chat)Answers questions about Canadian immigration programs in plain language, using your profile and official sourcesGive legal or immigration advice, submit applications, or contact government for youYou decide what to do; you can book a licensed RCIC or lawyer for advice
Program matchingCompares your profile with published criteria for federal, provincial and other programs and suggests ones to exploreDecide your eligibility or guarantee an invitation, nomination or approvalYou verify on official government sources
CRS and PNP calculatorsMostly rules-based calculations; AI may explain results and suggest ways to improve your scoreProduce an official scoreYou check results against official IRCC and provincial tools
Document assistantReads uploaded documents to extract details (for example, dates or test scores), flags possible gaps and suggests checklistsVerify that a document is genuine, or decide what you must declareYou review every extracted detail and form field before using it
Form pre-fill and draftingDrafts form answers, letters of explanation and summaries from your profileSign, certify or submit anythingYou must review, correct and approve every word
Job matchingRanks jobs for Users and candidates for Businesses based on skills, experience, location and preferencesMake hiring decisions, or use protected characteristics such as race, religion, sex or ageEmployers decide whom to contact and hire
Resume and profile helperSuggests improvements to resumes and profilesInvent experience or qualificationsYou approve all changes
Professional tools (RCICs and lawyers)Summarize client files, draft documents and research program rules inside the Professional’s workspaceReplace the Professional’s judgment or professional dutiesThe Professional is responsible for all advice and filings
Safety and fraud detectionFlags suspicious job postings, fake accounts, scams and policy violationsPermanently remove an account without human reviewOur trust and safety team reviews flagged cases
Customer supportAI may draft or answer support questionsHandle billing disputes or privacy requests without a personYou can ask for a human at any time
TranslationTranslates Platform content and AI answers into supported languagesProduce certified translations for government useCertified translators are required for official documents

3. AI Models and Providers

3.1 Models we use.

The Platform uses a combination of:

  • Third-party foundation models (large language models) accessed through enterprise APIs, from providers such as [insert providers, e.g. Anthropic, OpenAI, Google, Cohere]. We may choose a different model for each task based on quality, speed and security.
  • WorkSettle’s own models and components, such as rules engines for score calculators, search and retrieval systems that find official sources, matching and ranking models, and models we fine-tune for immigration-specific tasks (see section 5).

We publish the current list of AI providers, what they do, and where they process data on our AI subprocessor page at worksettle.com/ai-subprocessors, and we update it before adding a new provider.

3.2 What our AI providers may and may not do.

Our contracts with AI model providers require that they:

  • do not use WorkSettle data (your prompts, profile, documents or the AI’s outputs) to train, develop or improve their models;
  • use the data only to process our requests and return results to us;
  • retain data for no longer than [zero days ("zero data retention") / up to 30 days for abuse and safety monitoring only] and then delete it;
  • keep the data confidential and protected by appropriate security safeguards; and
  • tell us promptly about any security incident affecting our data.

If any provider’s terms differ from these standards, for example if a provider keeps data longer for safety review, we will say so on our AI subprocessor page and, where it involves sensitive features, ask for your consent first.

3.3 Where processing happens.

Our primary data storage is in Canada. Some AI providers process data in other countries, including the United States, as listed on our AI subprocessor page. See our Privacy Policy for more on cross-border transfers.

4. Data the AI Collects and Processes

When you use an AI feature, the following data may be processed. Section 5 explains separately whether any of it is used for training.

Data processed by AI features
DataExampleWhy the AI uses it
Your prompts and messages"Am I eligible for the BC PNP Skills Immigration stream?"To understand and answer your question
Profile contextAge, education, language scores, work experience, family details, current statusTo personalize answers and matches without asking you to repeat yourself
Uploaded documentsLanguage test results, passport bio page, diplomas, job offersTo extract details, check completeness and pre-fill forms, only when you ask
AI outputsAnswers, summaries, drafts, match results, score explanationsTo show you results, keep your history and let you continue later
Feedback signalsThumbs up or down, corrections, "report a problem"To find and fix errors
Usage and technical dataFeature used, time, response time, errors, device typeTo keep the service reliable, secure and fast
Safety signalsContent flagged as abusive, fraudulent or a prompt-injection attemptTo protect users and the Platform

What the AI does not do.

The AI Agent only uses the profile information and documents you have saved or shared in that session. It does not search the internet for information about you personally, buy data about you from data brokers, use your social media, or infer sensitive characteristics such as religion, ethnicity, sexual orientation or health from your data.

Information you should not share.

Do not enter passwords, banking details, social insurance numbers or other people’s private information in the AI chat unless a feature specifically asks for it. We apply automated filters to detect and mask some of this information, but they are not perfect.

Human access to AI conversations.

By default, WorkSettle staff do not read your AI conversations or documents. Limited, logged access by authorized staff is allowed only (a) when you ask us for help and give permission; (b) to investigate a safety, fraud or abuse flag; (c) to fix a technical problem; or (d) when required by law. Any quality review uses de-identified samples wherever possible.

5. Training Data: What We Use and What We Never Use

"Training" in this Policy means using data to build, fine-tune, evaluate or otherwise improve an AI model, including WorkSettle’s own models and any model a provider runs for us. Using your data to answer your own request is processing, not training.

5.1 Training data at a glance.

What is and is not used to train or improve AI
Data sourceUsed to train or improve AI?
Public government sources: IRCC, provincial nominee program, ESDC and Job Bank pages; the Immigration and Refugee Protection Act and regulations; published program guides, draw results and National Occupational Classification (NOC) dataYes. This is our main training and grounding data.
Licensed or purchased datasets (e.g. labour market data)Yes, within the licence terms. We do not buy data about individual people.
Synthetic data: made-up profiles and questions created by our team or by AIYes
Content written by WorkSettle’s team or contracted RCICs and lawyers to create test questions and expert answersYes, under contracts that assign or license that content to us and contain no real client information
Aggregated statistics (e.g. how many users ask about Express Entry)Yes, for product planning and evaluation; no individual can be identified
Feedback you give on an AI answer (thumbs up/down, a correction)Only in de-identified form: we remove names, contact details, document numbers and other direct identifiers.
Your identifiable AI conversations, profile and outputsNo, unless you turn on "Help improve WorkSettle AI" (opt-in, off by default)
Your uploaded documents (passports, permits, test results, diplomas, letters)Never, even if you opt in
Information about your spouse, dependants or referencesNever
Data in Professional workspaces (RCIC and lawyer client files)Never
Business data (candidate searches, internal hiring notes, private messages)Never
Private messages between Users, Businesses and ProfessionalsNever
Payment informationNever
Data from minorsNever

5.2 The "Help improve WorkSettle AI" setting.

If you choose to opt in, we may use your AI conversations and related profile fields to improve our models. Before any use, we:

  • remove direct identifiers such as names, contact details, addresses, dates of birth and document or application numbers, using automated tools and human checks on samples;
  • store training sets separately from production data, with restricted, logged access;
  • never use the data to identify you, contact you or make decisions about you; and
  • keep training sets only as long as needed [insert period, e.g. 24 months].

You can turn the setting off at any time. From then on, new conversations will not be used. We will also stop including your past conversations in future training runs and remove them from training sets we still hold. A model that has already been trained cannot "forget" its training. Because data is de-identified first, it is extremely unlikely that such a model contains information that identifies you.

5.3 No scraping of personal information.

We do not scrape social media, forums or other websites to collect personal information for training. Where we use public web content, we limit it to official government and other authoritative sources, respect website terms and robots.txt instructions, and exclude personal information.

5.4 Model evaluation.

Before release and after major updates, we test models against expert-written immigration questions, known edge cases and fairness tests. Evaluation uses the sources marked "Yes" above and never uses documents or Professional data.

5.5 If our practices change.

We will not start using any category marked "No" or "Never" above without updating this Policy, giving at least 30 days’ notice, and getting your express consent. A change will never apply to data collected before it, unless you agree.

6. Accuracy, Sources, Limitations and Human Oversight

6.1 Grounding in official sources.

The AI Agent is designed to base its answers on a library of official sources, such as IRCC and provincial program pages, legislation and published guides, which we monitor for changes [insert frequency, e.g. daily for draw results and weekly for program pages]. Where possible, answers show links to the sources used, along with the date the source was last checked.

6.2 Known limitations.

Even with these safeguards, AI can:

  • give answers that are wrong but sound confident (sometimes called "hallucinations");
  • miss a recent policy change, a new draw or a temporary public policy;
  • misread a document, especially a poor scan or a document in another language;
  • oversimplify complex situations such as past refusals, inadmissibility, criminality, medical issues or mixed status in a family; and
  • misunderstand a question that is unclear or missing key facts.

6.3 When the AI will refer you to a person.

The AI Agent is designed to recommend that you speak with a licensed RCIC or lawyer when your situation involves a past refusal or misrepresentation finding, inadmissibility, criminal or medical issues, loss of status, removal orders, refugee claims, appeals or judicial review, or when you ask for a specific recommendation on what to file.

6.4 Clear labelling.

AI-generated answers, drafts and summaries are marked as AI-generated. The AI Agent tells you it is an AI if you ask, and never claims to be a human, an RCIC, a lawyer or a government official.

6.5 Human oversight.

Qualified team members, including immigration professionals under contract, review our source library, test questions and model behaviour regularly. You can report a wrong or harmful answer using the "Report" button on any AI response or by emailing Info@worksettle.com. We review every report and correct systemic errors.

6.6 Your responsibility.

AI output is general information, not legal or immigration advice. You remain responsible for checking information against official sources and for everything you submit to any government authority, as set out in our Terms & Conditions.

7. Fairness, Job Matching and Automated Decisions

7.1 No automated decisions with legal effect.

Our AI does not make decisions that produce legal or similarly significant effects on you. It does not approve or refuse immigration applications, grant or deny jobs, or set prices for you personally. Those decisions are made by government authorities, Employers or you.

7.2 How job matching works.

Matching and ranking consider job-related factors only: skills, work experience, education, credentials, language ability, location and stated preferences, work authorization when an Employer requires it, and how recently a profile or job was updated. Matching does not use race, colour, ancestry, place of origin, religion, sex, sexual orientation, gender identity or expression, age, marital or family status, disability or any other ground protected by the BC Human Rights Code or the Canadian Human Rights Act. Nor does it use features designed to stand in for them, such as names or photos.

7.3 No pay-to-rank for job seekers.

A User’s subscription status never affects how their job application or profile is ranked or shown to Employers. Promoted job postings that Businesses pay for are labelled "Sponsored".

7.4 Bias testing.

Before launch and at least annually, we test matching, ranking and the AI Agent for unfair differences in results between groups, including by country of origin and language. We fix problems we find and record what we did.

7.5 Explanations and human review.

You can ask why a program or job was suggested to you. Where AI is used in a process that significantly affects you, such as an account suspension for suspected fraud, you can ask for a human to review it.

7.6 Employers’ responsibility.

Businesses that use AI-ranked candidate lists remain responsible for fair, lawful hiring decisions and must not use the Platform to screen candidates on protected grounds.

8. Professionals and Businesses: Confidentiality and Data Isolation

RCICs and lawyers owe their clients duties of confidentiality, and Businesses hold confidential hiring information. Our AI tools for them are built around those duties.

8.1 Professional workspaces.

  • Each Professional’s or firm’s client files, AI conversations and drafts are kept in a separate, access-controlled workspace. Other Professionals, Businesses and Users cannot see them, and the AI does not draw on one workspace to answer questions in another.
  • Professional workspace data is never used to train AI models, by WorkSettle or by our AI providers, and Professionals cannot opt in on their clients’ behalf.
  • WorkSettle staff do not access workspace content except at the Professional’s documented request for support, or as required by law.
  • Professionals can set retention periods and delete workspace data at any time. When a Professional leaves the Platform, we return or delete workspace data within 30 days of their request or contract end, unless the law requires us to keep it.
  • Professionals remain responsible for reviewing AI output, for their advice, and for meeting the rules of the College of Immigration and Citizenship Consultants or their law society, including any rules on disclosing AI use to clients or tribunals.

8.2 Business accounts.

Candidate searches, shortlists, internal notes and messages in a Business account are used only to provide services to that Business, are never used for training, and are never shared with other Businesses.

8.3 Enterprise terms.

Firms and Businesses may sign a Data Processing Agreement setting out these commitments in contract.

9. Security and Retention of AI Data

9.1 Security.

AI data is protected by the same safeguards as all personal information on the Platform: encryption in transit and at rest, role-based access controls, multi-factor authentication for staff, logging and monitoring, and vendor security reviews. We also defend against AI-specific risks, such as attempts to trick the AI into revealing another person’s data or ignoring its safety rules ("prompt injection"). We test these defences regularly.

9.2 Retention.

How long AI data is kept
AI dataHow long we keep it
AI chat historyUntil you delete it, or 24 months of inactivity
Uploaded documentsUntil you delete them or close your account
Copies held by AI model providers[None ("zero data retention") / up to 30 days for safety monitoring], as listed on our AI subprocessor page
Safety and abuse logs12 months, or longer if needed for an investigation or required by law
De-identified training sets (opt-in only)24 months
Professional workspace dataAs set by the Professional; deleted within 30 days of their request or contract end

10. Your Choices and Controls

In your account settings you can:

  • view, export and delete individual AI conversations or your whole AI history;
  • delete uploaded documents;
  • turn "Help improve WorkSettle AI" on or off (off by default);
  • choose whether the AI Agent may use your saved profile and documents, or only what you type in each chat;
  • turn off optional AI features, such as AI suggestions on your profile or resume, and still use the rest of the Platform;
  • ask why a program or job was suggested to you; and
  • request human review of any AI-assisted decision that significantly affects you.

You can also make access, correction and deletion requests as described in our Privacy Policy.

11. Governance and Accountability

  • Accountable executive. Our [AI Governance Lead / Privacy Officer] is responsible for this Policy and for approving new AI features.
  • Review before launch. Each new AI feature, and each new AI provider, goes through a privacy and risk review before it launches, covering the data it uses, its accuracy and fairness, and the risk of harm to users.
  • Frameworks we follow. We design our AI programme around Canadian privacy law (PIPEDA and BC’s Personal Information Protection Act), guidance from the Office of the Privacy Commissioner of Canada on generative AI, and the six principles of Canada’s Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems: accountability, safety, fairness and equity, transparency, human oversight and monitoring, and validity and robustness.
  • Incidents. If an AI incident causes or risks serious harm, for example a data exposure or a widespread error about a program rule, we will fix it, notify affected users and regulators as required by law, and publish a summary where appropriate.
  • Changing laws. Canada’s privacy and AI laws are evolving. We will update this Policy as new legislation comes into force.

12. Changes to This Policy

We will post updates here with a new "Last updated" date. For material changes, especially any change to how data is used for training, we will give at least 30 days’ notice by email or in-app message and ask for your consent where required.

13. Contact Us

Questions, concerns or reports about our AI can be sent to:

WorkSettle Inc. — AI & Privacy
7337 137 Street, Surrey, British Columbia V3W 1A4, Canada
Telephone: 604-341-3411
Email: Info@worksettle.com (subject line: “AI”)

If you are not satisfied with our response, you may contact the Office of the Privacy Commissioner of Canada or the Office of the Information and Privacy Commissioner for British Columbia.

Coming soon

Browse jobs in the WorkSettle app

Get matched to roles the day they are posted, apply in a few taps, and see where every application stands.

Download on the App Store (coming soon)Get it on Google Play (coming soon)