How we use AI, plainly explained

AI Policy

RF Holdings Corp., operating as “Gardril” · Calgary, Alberta, Canada
Last updated: October 18, 2026

Gardril sells regulated safety training. Our entire product promise is that the person named on a certificate is the person who actually did the training. A company that sells verified identity has no business being vague about its own use of artificial intelligence. This policy states plainly where we use AI, where we do not, and who is accountable for what you learn.

1. Why Gardril uses AI

We use AI so that regulated safety training costs workers less, without lowering the standard.

Producing a compliance course the traditional way means a studio, a camera crew, a hired presenter, a voice booth, and weeks of editing for every module and every revision. That cost is carried by the learner or by the employer, and it is the reason many courses have historically cost several times what they cost on Gardril.

Generative AI changes the economics of production, not the standard of the content. We use it to build and update courses faster and at a fraction of the cost, and we price our courses accordingly. It also means a regulatory change can be reflected in a course in days rather than in the next production cycle, which matters more in safety training than in almost any other subject.

What we do not do is use AI to replace the expert judgement that makes safety training worth taking. Every dollar AI saves in production is a dollar we do not have to recover from the learner. None of it is saved by skipping review.

2. Where AI is used, and where it is not

2.1 Course production

We use generative AI in the following ways, and we name them per course as described in section 3.

2.2 Identity verification and training integrity

Gardril uses Asgard Authenticate, wrapped inside the course itself, to confirm that the learner taking the course is the person enrolled. Asgard checks your identity before the course opens and again throughout it. If your participation drifts during the session, it prompts you and guides you back on track. It is also built to stop stand-ins and AI tools from completing the training on a learner's behalf.

This is AI used against fraud, not to teach. It is described in section 5, and the personal information it involves is described in our Privacy Policy.

2.3 Articles, resource pages and website copy

Gardril publishes articles and resource pages, and generative AI is used to help draft and edit that material and our website copy, in the same way it is used for course text. A person at Gardril writes, checks and approves everything published under the Gardril name. Where an article states a regulatory requirement, that requirement is checked against the primary source, meaning the regulation, the standard or the regulator's own published material.

2.4 What AI is never used for at Gardril

• AI is never the final authority on whether course content is correct. A human subject matter expert is, as set out in section 4.
• AI is never used to generate a testimonial, review, learner quote, endorsement or success story. Every review or testimonial attributed to a person is from a real person.
• AI is never used to imply endorsement by, or affiliation with, any government body or regulator. Gardril is not a government agency and no Gardril course is issued by one.
• We do not create a synthetic likeness or voice of a living person without that person's written consent covering that exact use. Where a video includes an AI recreation of a historical figure who has died, it is labelled as AI generated, it does not give that person a voice or put words in their mouth, and it never presents them as endorsing Gardril or any course.

2.5 Copyright and licensing

The tools and assets we use to build a Gardril course are paid for or licensed for commercial use. That includes the AI services we use for video, voice and imagery, and the stock libraries we use for photography, video and graphics. We use each one within the terms of its licence.

3. How every course declares its AI use

Every Gardril course carries an AI marking on its course page and within the course itself. There are two markings, and which one applies is decided by the rules that govern that course, not by our preference. The sparkle is the symbol used across the web to mean artificial intelligence, and the crossed sparkle means the opposite. Neither symbol is ever used on its own. The words next to it carry the meaning.

3.1 Marking A, for courses made with AI

3.2 Marking B, for courses made without generative AI

3.3 How a course is classified

• Gated courses, meaning any course where a regulator, ministry, board or standards body approves the course, the syllabus or the credential, are classified only against that body's written requirements. Where a governing body restricts or prohibits AI in course development, Marking B applies and the course is built to that standard from the start.
• Where a governing body's position on AI is not written down, we ask that body in writing and keep the answer. Until we have it, the course is either built to the Marking B standard or it is not sold as meeting that body's requirement. We do not guess, and we do not assume permission from silence.
• Non gated courses, meaning general awareness and employer directed training where no external body approves the content, carry Marking A where any part of the course was made with generative AI, and Marking B only where we hold the production record described below.
• Marking B is a factual claim about how a course was made. It means no generative AI was used to create the course content. It is applied only where we hold a production record proving it, including the software and settings used, and it is withdrawn immediately if any part of the course is later produced with generative AI. A course cannot carry Marking B because it sounds better.
• Classification is reviewed whenever a course is revised, whenever the governing body changes its requirements, and at least once a year.

4. Human accountability

AI is a tool. It is not a safety professional.

• Every published Gardril course is reviewed and approved by competent subject matter experts, and by qualified experts where regulations require it, before release. Gardril uses both internal subject matter expertise and contracted external experts, matched to the subject of the course.
• The review is documented. We record who reviewed a course, against what standard, and on what date, and we keep that record for as long as the course is sold and afterwards.
• Review specifically checks for the failure modes generative AI is prone to: factual error, invented or misquoted regulatory requirements, out of date legislative references, and content that is technically accurate but unsafe as practical guidance.
• Regulatory content receives the closest scrutiny, because it is where an AI error would do the most harm. Where a course states a legal requirement, that requirement is checked against the primary source, meaning the regulation, the standard or the regulator's own published material, and never against a secondary summary.
• Content accuracy and content integrity remain human responsibilities of Gardril. No disclaimer transfers that to a vendor, to a model, or to you.

We also do not claim regulatory status we do not hold. Where a course is not approved, authorized or accredited by a particular body, we do not say or imply that it is, and where a regulator requires a live qualified instructor or a specific approved delivery format, a self paced Gardril course does not substitute for it and is not sold as doing so.

5. AI in identity verification and training integrity

Our checks guide you through. They do not decide your result.

5.1 What happens

Asgard checks your identity before your course opens and again during it. If your participation drifts, it prompts you and guides you back on track, and the session continues. A check can stop you from continuing, for example if your identity cannot be confirmed or someone else appears to be taking the course for you. AI does not mark your assessment and does not decide whether you pass. Each check is recorded as part of your completion record.

5.2 If a check stops you

• We will tell you that an automated check stopped you.
• On request we will tell you what personal information was used to reach it, the reasons and the principal factors behind it, and how to have that information corrected if it is wrong.
• You can require a human review. You may send us your account of what happened and a member of Gardril personnel with authority to reverse the outcome will consider it and respond. Write to privacy@gardril.com with the course name and the date.
• No credential is revoked by an automated check alone. A revocation is a human decision, made by a named person, on reviewable evidence.

This is the standard set by Quebec's private sector privacy law for decisions based exclusively on automated processing, and we apply it to every learner in Canada and the United States rather than only where a statute compels it.

5.3 What we do not claim about it

We describe what our integrity technology does. We do not claim it is infallible. Terms like cheat proof, tamper proof, unspoofable or guaranteed do not appear in our material, because no such system exists.

6. Jurisdiction specific disclosures

Gardril applies one standard everywhere, and adds what a particular jurisdiction requires on top.

6.1 New York

Where a Gardril advertisement contains a synthetic performer, the advertisement conspicuously discloses it, as required by New York General Business Law section 396-b. Gardril's websites and promotional videos may also show AI generated people. Where they do, the page or video says so where the person appears, the people shown are not real Gardril employees, learners or customers, and none of them is presented as giving a review or testimonial. The synthetic presenters described in section 2.1 also appear on our course landing pages. There, as in our advertising, the image or video carries a visible made with AI stamp.

6.2 California

California's AI transparency regime places provenance and disclosure duties on the large providers of generative AI systems, not on the companies that use their tools. Gardril's obligation as a customer of those providers is to leave intact the provenance information embedded in what they generate for us. We do not strip it during course packaging.

6.3 Anywhere we deploy a chatbot or conversational assistant

Any Gardril conversational tool states at the outset, and while you are using it, that you are interacting with AI and not a person. This is required in several states and is our practice everywhere. Gardril does not operate a companion or emotional support chatbot, and no Gardril AI tool offers medical, legal, financial or mental health advice.

6.4 Canada, federally

Gardril aligns its practice to the federal Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems as a manager of AI systems, and to the joint privacy commissioners' principles for generative AI, including the principle that outputs which significantly affect a person are identified as AI generated and that people know when they are dealing with an AI tool.

7. Limitations we will state plainly

• Generative AI produces confident errors, including invented citations and misstated legal requirements. This is the reason for section 4, not a residual risk we hope you will not notice.
• AI review is risk tiered, not uniform. Everything that bears on a credential, a regulatory requirement or a safe work procedure gets full subject matter expert sign off. Supporting and enrichment material is reviewed to a proportionate standard. Promising that a human reads every single AI output at every stage would be a promise no training company can keep, so we do not make it.

8. Governance and changes to this policy

This policy is owned by Gardril's Chief Executive Officer, who is accountable for its content and for the AI decisions described in it. It is reviewed at least annually and whenever any of the following happens: a new course category is launched, a new AI tool, stock library or other content vendor enters course production, a vendor changes its licence terms, the identity verification or credential path changes, or a law in a jurisdiction we sell into changes what we must disclose. It is important to note that AI governance and regulations are changing constantly. Gardril makes its best effort to meet its obligations as required.

9. Frequently asked questions

Does Gardril use AI to create its courses?

Yes, in some courses, and every course says which parts. We use generative AI for video presenters, narration, imagery, draft text and, in some courses, draft assessment questions. Competent subject matter experts, and qualified experts where regulations require it, review and approve every course before it is published, and Gardril remains accountable for accuracy. Courses governed by a body that does not permit AI in course development are built without it and carry the made without generative AI marking.

Is the person presenting my course a real instructor?

In courses marked made with AI, no. The presenter is a synthetic avatar with synthetic narration, and the course page and the course itself say so. The presenter is never given a professional designation or credential, because it does not hold one. The expertise is in the content and in the named human expert who reviewed it.

Did a human check this course for accuracy?

Yes. Every published course is reviewed and approved by competent subject matter experts, and by qualified experts where regulations require it, before release, using internal and contracted expertise matched to the subject, and the review is documented with the reviewer, the standard and the date. Regulatory statements are checked against the primary source rather than a summary.

Who owns the copyright in AI generated course material?

Gardril builds courses with paid, commercially licensed tools and stock libraries, and uses each one within its licence. The substance of every course is written and approved by Gardril and its subject matter experts.

Which US states require an AI policy for online training?

No US state currently requires an online safety training provider to label a course as containing AI generated content. What US law does require is that you not be misled: federal and state consumer protection law prohibits creating the impression that a human expert taught a course when none did, using synthetic people as endorsers, implying government affiliation, or overstating what an anti fraud technology achieves. New York separately requires conspicuous disclosure of a synthetic performer in advertising. Gardril publishes this policy because the honest version of the answer is better for learners than the minimum, and because our own product is sold on trust.

10. Contact

Privacy and automated decision requests: privacy@gardril.com
General support: support@gardril.com
RF Holdings Corp., operating as Gardril · Calgary, Alberta, Canada