AI tools in recruiting: what the law requires of employers
Monday, 7:30 a.m. Your building maintenance job ad drew a pile of applications over the weekend, and your new tool promises to sort them in two minutes. Tempting. Then, on Wednesday, a rejected applicant writes: "Why not me?" If nobody looked at their file, Québec law expects you to have an answer. Here's what it requires, and how to apply it without pulling your hair out.
Where AI is already slipping into your hiring
You don't need to have bought "AI software" to be affected. Artificial intelligence often hides in everyday features: automatic sorting of applications, a match score, a chatbot that asks screening questions, a tool that writes your job ad or suggests interview times.
The law doesn't care about the brand of the tool. It cares about what the tool does with applicants' personal information, and above all about one question: is a machine making a decision about a person on its own? For the broader picture of what AI is changing in the service trades, read Will AI replace the service trades? What is really changing.
- AI helps you: it drafts a job ad, summarizes applications, suggests a reading order. You read and you decide.
- AI decides: it rejects applications or closes files without a person looking at them.
Do it today: list every tool that touches your applications (form, sorting, messaging, writing) and write next to each one: "helps" or "decides."
The core rule: section 12.1 of Law 25
Law 25 amended Québec's Act respecting the protection of personal information in the private sector. Most of its provisions have been in force since September 22, 2023, including section 12.1 on automated decisions. It applies when you use personal information to make a decision based exclusively on automated processing.
In that case, you must:
- Inform the person that the decision was made through automated processing, no later than when you inform them of the decision.
- On request, tell them what personal information was used to make the decision, the reasons and the main factors and parameters that led to it, and that they have the right to have that information corrected.
- Give them the opportunity to submit observations to a member of your staff who is in a position to review the decision.
In practice, Wednesday's rejected applicant has the right to know a machine screened them out, on what basis, and to speak with a human who can change the decision. Not another bot.
Do it today: name the person on your team who will review automated decisions, and prepare a short template message telling applicants how to reach them.
"Exclusively automated": does it apply to you?
The key word is "exclusively." If a human really looks at the file before deciding, section 12.1 doesn't cover that decision. But watch out for fake human oversight: clicking "approve" on 80 rejections at once without reading them looks a lot like a machine decision.
| Situation | Exclusively automated decision? |
|---|---|
| The tool ranks applications, you read all of them before choosing | Generally no: you're the one deciding |
| The tool automatically rejects anyone who didn't check "Class 3 licence" | Probably yes, if nobody reviews those files |
| A screening chatbot closes the file after a wrong answer | Probably yes |
| AI writes your job ad | No, there's no decision about a person |
Borderline cases exist. When in doubt, treat the decision as automated: informing people and offering a human review costs you almost nothing. For a specific case, the Commission d'accès à l'information (CAI), Québec's access to information commission, or a lawyer can help.
Do it today: for every tool marked "decides," check its settings to see whether it rejects applications without human review. If it does, add a review step or apply section 12.1.
The other rules that apply to your tools
Section 12.1 isn't alone. An AI tool is still subject to all the usual rules on personal information and discrimination.
- Collect only what's necessary (s. 5): only the information needed for the purposes you set before collecting it. A form asking for a date of birth for a window washer position? No.
- Profiling (s. 8.1): if you collect information using technology that can identify, locate or profile a person, you must inform them beforehand, along with the means available to activate those functions.
- A new system (s. 3.3): a project to acquire or develop an information system involving personal information requires a privacy impact assessment, proportionate to the sensitivity of the information.
- A person in charge (s. 3.1): in your business, the person with the highest authority is responsible for protecting personal information; they can delegate that role in writing.
- Québec's Charter of human rights and freedoms: no discrimination in hiring (s. 16), and no questions on the grounds listed in section 10 (age, sex, pregnancy, origin, civil status, disability…) in an application form or interview, except as provided (s. 18.1). A tool that screens people out on those grounds, even unintentionally, is a problem: the effect is what counts.
The penalties aren't symbolic: according to Québec.ca, the CAI can impose administrative monetary penalties of up to $10 million or 2% of worldwide turnover, and penal fines can reach $25 million or 4%. Also hiring in Ontario with 25 or more employees? Since January 1, 2026, a publicly advertised job posting must say whether AI is used to screen, assess or select applicants.
Do it today: reread your application form and strike out every question that doesn't help you decide whether the person can do the job. Your job ad follows the same rules: Writing a job post that attracts candidates (and follows the law).
How GPP Emploi applies these rules
We built the platform so employers don't have to guess. Matching is based on a match score out of 100, always explained: trade 35, location 20, availability 15, cards 15, pay 10, working languages 5. Nothing else is used: not age, not sex, not origin.
The score ranks; it doesn't reject anyone. In the profile bank, you see an anonymous profile (trades, region, availability, cards) and you're the one who sends the invitation. The worker accepts or declines, and if they decline, you'll never know who they are. The hiring decision stays yours, from start to finish.
When the system does make a decision, for example holding back a job ad before publication or restricting an account, the person sees why and can ask for a member of our team to review the decision, with a response within 5 business days. In "Mes données" (my data), workers can download their data or delete their profile. It's all detailed on how it works and in the privacy policy.
Do it today: read the employer page and compare explained matching with what your current tool shows you. If you can't explain a rejection, neither can your applicant.
Your checklist before using an AI tool
You don't need a legal department to start off right. Half an hour with this list covers the essentials, whether you're hiring a snow removal worker, a house cleaner or an apprentice painter.
Do:
- Know what information the tool uses, and why each piece is necessary.
- Be able to explain in two sentences why an applicant was screened out.
- Have a person who rereads automatic rejections and answers review requests.
- Tell applicants, starting on the form, that an automated tool is involved.
- Use the same grid for everyone.
Avoid:
- Filters on age, career "gaps" or exact address, which can screen people out for the wrong reasons.
- Pasting full applications into a consumer tool without knowing where the data goes.
- Letting the tool send rejections without review. A bot saying no at 3 a.m. is efficient, but not very warm.
For AI uses that really help you day to day, without deciding for you, read AI for small service businesses: 10 practical uses starting tomorrow.
Do it today: print this list, check off what's already in place and pick one empty box to fix this week.
The bottom line
AI can sort, summarize and suggest; a decision about a person must stay explainable and reviewable by a human. If you can explain every rejection in two sentences and name who reviews it, you're already in good shape.
Frequently asked questions
Can an employer use AI to screen applications in Quebec?
Yes: Law 25 doesn't ban these tools, it regulates automated decisions. If a decision is based exclusively on automated processing, the employer must inform the applicant, explain on request the information used and the main factors, and let them submit observations to a person who can review the decision.
What is a decision based exclusively on automated processing?
It's a decision made by a system with no real human involvement, such as an automatic rejection of an application. If a person actually reviews the file before deciding, the decision isn't exclusively automated. For a specific case, check with the Commission d'accès à l'information.
What is the penalty for not complying with Law 25?
According to Québec.ca, the Commission d'accès à l'information can impose administrative monetary penalties on businesses of up to $10 million or 2% of worldwide turnover. Penal sanctions can reach $25 million or 4% of worldwide turnover.
Do you have to say in a job ad that you use AI?
In Québec, the law requires you to inform the person of an automated decision no later than when you communicate the decision. In Ontario, since January 1, 2026, employers with 25 or more employees must say so in publicly advertised job postings. Mentioning it in the ad is good practice everywhere.
Can an AI tool discriminate against applicants?
Yes, if it was set up or trained in a way that screens people out on a ground protected by the Charter, such as age, sex or origin. The Charter looks at the effect of a distinction, not just the intent. The employer remains responsible for the tool it chooses to use.
Sources
- P-39.1 - Loi sur la protection des renseignements personnels dans le secteur privé
- C-12 - Charte des droits et libertés de la personne
- Loi 25 sur la protection des renseignements personnels des citoyens du Québec - Entrée en vigueur de nouvelles dispositions qui font du Québec un chef de file mondial
- Requirements related to publicly advertised job postings
