For software developers in the Talent Tech industry, Artificial Intelligence (AI) has been a game changer, speeding up workflows and freeing up time and resources. But when looking to work with an AI provider, what are the key issues developers should be alert to before signing on the dotted line? Here Chris Mooney, Partner and Liberty Pearl, Associate, outline some of the key contractual issues that developers may encounter when looking to use AI tools for coding.
Developers across the Talent Tech industry have been quick to take up AI tools to help drive efficiency in their day-to-day work. The benefits AI brings, in maximising efficiencies and minimising potential human error are obvious and present an attractive option for businesses looking to maximise their resources. That said, there are also legal issues that need to be considered before using AI tools for coding.
Five Key Issues
-
Restrictions on the use of an AI tool
The terms and conditions governing the use of AI tools (“AI T&Cs”) typically include provisions that set out how the AI tool may (and may not) be used. Businesses must ensure that they comply with these restrictions to avoid breaching the contract with the AI tool provider and facing potential legal disputes or losing the right to use the AI tool in the future.
Notably, the AI T&Cs will almost certainly make the business responsible for the acts and omissions of its coders (employees or contractors) when they use the AI tool. Therefore, the business will need to put a staff policy in place to flow these restrictions down to its coders and make them aware of the restrictions on their use of the AI tool.
Action: The AI T&Cs should be checked carefully. Make sure they do not prevent the business from using the AI tool for its intended purpose and also check and / or amend your staff policies to ensure coders adhere to any restrictions.
-
Intellectual property (“IP”) ownership and licensing
There are two important areas to consider – input ownership and licensing and output ownership and licensing.
The general position adopted by the AI T&Cs on inputs is that the business using the AI tool retains ownership of the IP in its inputs and grants the AI tool provider a licence to use them.
However, the scope of the licence granted to the AI tool provider can vary significantly. It’s important businesses review the scope of the licence they are granting the AI tool provider to make sure it is not too wide, particularly where the inputs are confidential or commercially sensitive.
The position on outputs is more complex because it is an area of law that is developing around the world.
In the UK, it appears that copyright protection can subsist in the outputs created by AI tools and that the owner of the copyright will be the person that made the arrangements necessary for the output’s creation, which will be either the AI tool provider, the business or the individual user.
Businesses need to consider the contractual matrix to ensure they are in the best position to acquire ownership of the IP in the code they create using AI tools.
Also, businesses should make sure their contracts with their coders include an active assignment provision, so that if any IP in the code vests in the coder as the individual user of the AI tool, it is automatically assigned to the business on creation.
Finally, it is important to consider whether the AI T&Cs grant the AI tool provider a licence to use the outputs and the scope of the licence. For example, this could be an issue if the AI T&Cs grant the AI tool provider the right to use the generated code to further train the AI model, particularly if the code is confidential or commercially sensitive.
Action: Getting IP provisions right is essential. Consider getting good legal advice before signing your contract.
-
Potential third-party IP claims
If businesses use AI tools for software development, there is a risk the generated code could infringe the IP in existing third-party code, which could lead to third-party IP infringement claims being made against the business.
This risk arises due to the fact AI tools are trained on large datasets and will typically reproduce the training data (or versions of the training data) as part of the outputs. If the AI tool has been trained on third-party software code without the correct permissions, and then reproduces some of the third-party code as part of the output, the owner of the third-party code could bring an IP infringement claim against the business when it starts using the output.
The risk of third-party IP infringement claims is a key issue for businesses using AI tools for software development.
In our experience, the AI T&Cs are often defective when it comes to protection from third-party IP infringement claims. They will either say the outputs are provided on an “as is” basis without protection or, where an indemnity is given, the AI tool provider’s liability is capped to such an extent that the indemnity is effectively worthless.
Action: Ensure the AI T&Cs give you adequate protection from third-party IP claims. Otherwise your business will need to consider other risk mitigation strategies, such as only using AI generated code for internal business purposes or limiting the amount of AI generated code that is incorporated into final products.
-
Confidentiality
Businesses should make sure the AI T&Cs contain robust confidentiality provisions to protect the business’ inputs and outputs within the AI tool.
If the confidentiality provisions are defective, there is a risk that the business’s inputs and outputs could be leaked to third parties such as other users of the AI tool.
Action: Carefully check the AI tool confidentiality provisions to ensure your inputs and outputs are protected.
-
Open-source software
Many AI tools have been trained on open-source software, which can lead to complicated issues around licensing and the use of AI generated code.
Open-source software code is made available for anyone to use free of charge on either permissive or restrictive licence terms. Where it is licensed on permissive terms, it can be freely used without restriction. However, where it is licensed on aggressive restrictive terms, it can impose requirements that any software which incorporates or links to the open-source software must also be made available free of charge on the same open-source basis.
This can cause significant issues for businesses when their own proprietary code has been built on open-source software which is licensed on aggressive restrictive terms.
The issue with AI tools is that if the AI tool has been trained on open-source software which is licensed on aggressive restrictive terms, and that open-source software is reproduced in the generated code, then the generated code, and also any other software code which incorporates or links to the generated code, could be required to be made available free of charge on the same open-source basis.
This will be a significant issue where the code is one of the business’s key proprietary assets.
AI T&Cs do not usually give any assurances on the use of open-source software by AI tools, so there is always the risk that open-source software which is subject to aggressive restrictive terms could be included in the generated code.
Action: Consider the need for other risk mitigation measures, including reviewing the AI tool T&Cs to see if it uses open-source software code, and checking and scanning the generated code for such open-source software code.
AI offers massive opportunities for software developers in the Talent Tech industry. However, if thinking about using AI, developers need to be vigilant on the details of the contract they will enter into by using the AI tool. Those already using AI might well want to review their existing contracts to make sure they’re best-fit for their business needs.
We are regularly assisting clients with their AI projects. If you would like advice on any aspect of AI, please feel free to get in touch; Chris Mooney Liberty Pearl
