AI and software: Crafting cutting-edge contracts for the digital frontier

Welcome to the first blogpost in our new series “AI & IP,” in which we’ll guide you through the impact of AI on intellectual property and technology law in a practical, hands-on manner. Given the omnipresence of AI news nowadays, we assume that an extensive introduction to AI isn’t necessary. Simply speaking and to freshen up your mind, AI is a technology that imitates human cognitive intelligence such as machine learning and problem-solving, posing multiple different legal challenges. The first topic to discuss in this series is AI in the context of software contracting.

 

Standard software contracts

In traditional software contracts, where suppliers develop or supply software solutions (be it as a service or as a product) to their customers, both the supplier and the customer must consider several important aspects when negotiating their contracts. Think about the typical hurdles like clauses concerning exclusivity, liability, warranties, service levels, IP and third-party software, personnel, customer data, security, escrow and indemnities.

 

AI contracts

Developing or supplying AI tools as such

With the rise and further expansion of AI, new contracts have made their entrance, such as AI-tool supply contracts, AI-tools-as-a-service contracts and AI-tool development contracts. These new contracts give rise to (already) commonly identified AI challenges, urging suppliers and customers to amend their standard clauses as used in their standard software contracts and to include new clauses. Examples include clauses on the AI tool’s training model (particularly on which data is being used and how), errors (suppliers may want to stipulate that the AI tool is provided on an “as-is” basis and that the AI tool will therefore not be uninterrupted or error free), measures to mitigate bias risks and related liability, customer data (customers may want to include a prohibition to use customer data in training processes), record keeping and audits of the AI tool.

That said, the issues arising in the field of IP law and the negotiations involved may be even more complex. Specific examples are:

• Ownership: A clause may need to be included on the ownership of the AI tool, the ownership of (or access to) the underlying software/IP and the ownership of the generated output.
  
  
• Third-party IP: Customers may want to include a warranty that the AI training processes or generated outputs do not infringe any third-party IP rights and/or, as a softer commitment, require the supplier to incorporate technical measures or tools to avoid the AI training processes or the generated output infringing third-party IP rights.
  
  
• Open source software: AI tools may be trained on open source libraries, including copyleft software, meaning the generated output could be subject to the same copyleft license terms. Appropriate clauses may need to be included to address this and mitigate any risks involved (eg the obligation for the customer to make the generated output publicly available).
  
  
• Indemnification: Customers may want an indemnification obligation for the supplier in case of third-party IP infringement by the AI training processes (data mining) or by the generated output.
  
  
• Usage rights and restrictions: A clause will need to be included on usage restrictions and other license terms of the AI tool itself or of the generated output. The supplier may want to limit the use of generated output to internal use only (and thus prohibit any further commercialisation).

The specific implementation of any of these clauses will of course be subject to negotiations between the supplier and the customer and depend on the exact nature and subject matter of the contract, on applicable laws and on the priorities, business model and business strategy of the contracting parties.

Use of AI tools to develop software

AI contracting may be even more challenging where the supplier does not develop or supply AI tools “as such,” but rather uses AI tools to develop software, which is then supplied to customers. In this case, a lot depends on the contractual terms of the AI tool, as the supplier may not want to make any commitments vis-à-vis its customer that go beyond these contractual terms.

For example, this will be the case in the absence of an indemnification obligation for the provider of the AI tool in the event of third-party IP infringement by the generated output. Conversely, even where the contractual terms of the AI tool do contain an indemnification clause, the question would still be to which extent the indemnification obligation would apply if the generated output is further modified by the supplier during the software development process.

The same goes for the ownership of the generated output. If the AI tool disclaims ownership to any generated output, but does not warrant that the generated output will not infringe any third-party IP rights, the warranties may of course also be absent in the contract between the supplier and the customer, following which the question of ownership of the final software could remain uncertain.

As a last example, the aspect of open source software may pose further challenges when a supplier develops software for its customer while using AI. In standard software contracts the supplier can control which open source software is being used in the software development process (and limit any risks regarding copyleft license terms). But it may be more complex for the supplier to provide any certainties in this respect if the AI tool used to develop the software operates on a black box (meaning that the training process and data are not known to the user of the AI tool, which will often be the case).

 

Building an AI mindset

The rise of AI contracts entails many contractual challenges, including in the field of IP. This will not only force both suppliers and customers to thoroughly review their current standard software contracts, but also to come up with creative, out-of-the-box solutions, where both parties will inevitably have to compromise. One thing is clear: adopting an AI mindset for both supplier and customer is key to benefit from the opportunities that AI contracting brings. Please reach out to your DLA Piper contact should you need further advice or assistance with respect to any of the above.