For certain borrowers, particularly those in the technology and creative sectors, intellectual property (“IP”) can be among their most valuable assets. Traditionally, lenders have not always performed significant due diligence on a borrower’s IP assets, but the rising use of artificial intelligence (“AI“) is yet another reason why that approach may deserve a closer look.

AI tools are increasingly being used by companies and their employees to generate a variety of work product results including software code, written and visual content, and even branding materials. This evolution raises several practical questions for secured lenders – How confident can you be that any IP rights can be claimed in the work product results created with the assistance of AI, or that the borrower actually owns such results? As AI requires large quantities of data collected from third parties, how reliable are rights in AI results in light of risks of IP infringement claims or IP invalidation? These issues raise uncertainties as to whether the results  that your security interest will attach to will be enforceable against the relevant collateral.  

What hasn’t changed

The personal property security legislation in each of the provinces and territories remains the starting point for a lender taking security over IP and the perfection of such security interest. IP, whether generated with the assistance of AI or otherwise, including copyrights, patents, trademarks, industrial designs and their related licences, generally fall within the definition of “intangibles” under personal property security legislation and perfection is achieved by registration based on the debtor’s location (the jurisdiction of incorporation or chief executive office, depending on the applicable province or territory).

While not strictly required for perfection, many lenders also record notice of their security interest in registered trademarks, patents, industrial designs, and copyrights at the applicable offices within the Canadian Intellectual Property Office (“CIPO”). Recordation of a security interest at CIPO alone does not perfect such interest, but rather provides public notice that a charge exists.

New due diligence considerations in the AI era

The growing use of AI means that traditional IP due diligence may need to go further than relying on just a registry search and the traditional representations and warranties contained in lending and security agreements to the effect that a borrower “owns” its IP. The concern being not that the method of perfection has changed but that the borrower may not actually have sufficient rights in the collateral to allow perfection to occur.

The following are examples of new due diligence questions that a lender may want to consider as part of its underwriting process:

1. Who actually owns the IP?

The use of generative AI tools in the creation of a work product may complicate ownership of IP rights therein. Even if the contract governing use of the generative AI tool states that the user will own the work product, can the AI tool pass ownership of the work product to the user? To do so, the AI must have rights of ownership in the work product. Many current Canadian IP laws grant first ownership of IP to a human creator, author, or inventor. Non-human AI therefore may not be an owner of any IP at law. Consequently, neither the AI nor the user may own IP rights in any AI work product.

Thus, if a borrower’s IP assets were produced with the assistance of generative AI tools, ownership may not be as straightforward as in the past and the borrower may have cloudy (or no) ownership rights in the IP collateral.

Some steps that a lender may want to take before advancing funds include:

  • Asking whether AI tools were used in creating any material IP;
  • reviewing the agreements governing the use of any AI tools to determine whether the agreement assigns to the user ownership of all IP and other work product resulting from the user’s use of the AI;
  • enquiring about the source of the data in the datasets utilized by the AI tool, and use of the data is authorized and lawful for the AI and user; and
  • ensuring the borrower has maintained records of human contribution where AI assistance was used.

These steps can be very important because if the borrower does not have sufficient ownership in the collateral, the lender’s security interest may not attach to such collateral, which limits the lender’s enforcement rights.

2. Are AI generated work assets protectable as IP?

Although applicable law in Canada continues to evolve, works generated by generative AI may not qualify for IP protection in Canada. Under current Canadian laws, copyright and design rights are typically granted to a human author or designer (as are some other IP rights). Presently there is no settled answer as to how to apply existing IP laws to work product produced by a generative AI tool. On a strict application of such Canadian laws, work product produced by the non-human AI may not be granted any IP rights.

Issues of IP rights and ownership are presently under review in many countries, and to date, the results of those reviews have been inconsistent. Some countries require IP rights to solely be granted to humans, and other jurisdictions have determined that a user providing a sufficiently erudite instruction can be granted IP rights in the work product resulting from use of the AI. As there is no consensus internationally at this time on these issues, for borrowers holding multi-national IP portfolios in particular, the variances in IP rights and ownership are critical considerations.

Uncertainty as to whether or not particular collateral can be protected may affect the value of such collateral, and should be discussed with counsel during the diligence process.

3. Potential IP infringement allegations

It is a well known fact that AI requires large datasets to be trained and to operate. If the data is obtained and used without authorization (e.g., via data scraping, etc.), this can cause the IP rights in AI generated work product to be at risk and consequently unreliable. Such unauthorized use can result in IP infringement and other rights challenges. A successful IP infringement challenge can lead to the loss of IP rights. Therefore, the AI datasets can be the basis for some uncertainty regarding the reliability of AI generated IP as a collateral.

4. Can the lender realize on the IP if it needs to?

Even if perfected, enforcement over AI-generated IP can be complicated if a lender can’t practically access, operate, or transfer the IP on enforcement.

In order to protect themselves, lenders should verify:

  • whether the security interest expressly extends to proceeds – in many cases, the value of IP lies in its proceeds;
  • the location of the assets – if they are stored at the premises of the borrower, in the cloud, or on hosted platforms;          
  • whether there are access credentials and keys available; and
  • whether the borrower’s agreements with third-parties permit transfer rights on default.

Without these safeguards, a perfected security interest may exist, but may not be realizable in any enforcement proceedings.

Strengthening representations, warranties and covenants

Given these uncertainties, lenders may wish to enhance standard IP representations and covenants in their loan and security documents to address AI-related risks.

Examples include:

  • Ownership and authorship representation:
    “The borrower owns, or has valid and enforceable rights to use and grant a security interest in, all IP forming part of the collateral and no portion of the IP was created or generated, whether through the use of generative AI or otherwise, in a manner that would fail to generate, or would negate, invalidate or diminish ownership, rights or protection under Canadian law. The collection and use of all data and datasets of AI tools used by borrower was authorized and lawful.”
  • Compliance with licences:
    “The borrower’s use of AI tools, models, and datasets has complied, complies and will comply with all applicable licence terms and AI, data protection and privacy laws, regulations and guidelines governing such use, and such use does not restrict the grant of a security interest in favour of the lender or the transfer of the IP or other results of such use in connection with any enforcement proceedings commenced by the lender.”
  • Maintenance covenant:
    “The borrower will maintain IP registrations in good standing and will pay all maintenance, renewal and other fees relating thereto and submit all filings required therefor, and borrower will maintain valid and enforceable rights in, and preserve evidence of authorship and ownership for, works developed through the use of any AI and all IP relating thereto.”

These provisions will not change how the security interest is perfected, but instead they are in place to help ensure that the security interest will attach in to collateral in a manner that is enforceable and transferable.

Bringing it together

Overall, the use of AI by borrowers has not required a fundamental change in approach by lenders. Registration under applicable personal property security legislation is still required and filing remains the key to perfection.

What has changed is the factual and legal certainty around a borrower’s ownership and control of underlying IP when AI is used in the business of the borrower. Lenders must now look deeper at how the results of the AI use were created, whether any IP was generated, under what terms the AI is used, and whether any ownership of IP or other AI results are in fact assignable in an enforcement proceeding.

By combining traditional secured lending principles with an updated due diligence framework, coupled with the addition of specifically tailored representations and covenants, lenders can continue to fund as confidently as possible, even as the use of AI increasingly affects the assets they secure.

If you’re taking security over IP in an AI‑driven business, now is the time to look a little deeper. Our Financial Services lawyers and Intellectual Property lawyers work closely with lenders to stress‑test IP and other results collateral generated through use of AI, refine due diligence, and tailor security documents to reflect emerging AI risks, so you can fund with confidence and protect your position.