AI and construction contracts: communication is key

Risk can usually be mitigated in construction contracts, but the fast-moving development of AI (with growing opportunities for deployment) gives rise to legal uncertainty, with many grey areas remaining unresolved in law or the courts so far.

So what can construction parties consider to protect themselves at this early stage?

At the moment, AI is developing and being used in many construction stages, including:

While these developments are exciting, parties deploying these tools need to consider important issues arising from AI use.   

Where will AI be used and what for?

It’s important to have clear discussions with the right people to establish whether AI will be used and for what purpose and, equally important, to agree whether there are circumstances in which it should not (or cannot) be used. This can be recorded in your construction contract if it’s important.

Is input information confidential?

Do you know whether the AI tool being used is an open or a closed model? Open models often give the AI provider the right to access the information put into the model and use it (eg to train the AI tool further), with the risk of that information becoming public. Closed models tend to have more protections, depending on the terms of the licence to use the AI.

From a construction contract perspective, consider:

Furthermore, if any parties involved in construction contracts use shared information to develop their own AI tools, they should be aware of the EU AI Act’s requirements for general-purpose AI providers.

Providers that fall within the scope of the Act must have a policy explaining how their models comply with existing EU copyright law and must publish a summary of the training data used to develop those models. As a result, information shared in the context of construction contracts and later used for AI tool development could fall within these reporting requirements.

Clearly, parties should discuss these topics and understand the nature of the information (and how proprietary and/or confidential it is), how and which AI tools will be used (open or closed, under what AI licence terms) or developed, and record any agreed terms in the contract.

Who owns the AI output?

Does the party that provides the AI tool, or the party inputting the prompts into the AI tool, own the output from that tool? UK copyright law stipulates that the author of a “computer-generated work” is the person who takes the arrangements necessary for the creation of the work – but it’s unclear whether this means the AI tool provider or user.

As this legal position is grey, it may come down to the licensing terms for the AI tool. Indeed, if the AI provider is an external party that retains ownership of the output (but grants a licence for use), this may impact the negotiation of your construction contract, and you must ensure it is reflected back-to-back down the contractual chain, eg with subcontractors.

Is the AI output protected by existing intellectual property laws?

This is another tricky question, the answer to which is unclear at the moment. Broadly speaking, UK intellectual property laws focus on protecting and incentivising human endeavour, innovation and creativity (trade mark law aside).

Whether or not AI output is ever truly “novel” or “original” remains an outstanding question for the courts to consider in the future. As such, copyright clauses in your construction contract might need more scrutiny where AI tools will be used in the delivery, and contractual restrictions on use of such outputs may need to be tightened if parties need to rely on breach of contract remedies (where proving ownership of intellectual property for infringement claims proves more difficult).

Who is responsible for errors in AI output?

If AI gets it wrong (eg materials run out or don’t arrive on time, the design or model has material errors, or face recognition ID lets unauthorised personnel on site), who is responsible? Although AI is billed as autonomous, if something goes awry, parties will look for someone to blame – and AI has no legal personality, nor assets against which to make a claim.

Practically speaking, consider what you are willing to take responsibility for if you choose to deploy AI to deliver a service (as you would an employee or subcontractor), and think about the negotiated position with all other relevant parties – those party to the construction contracts as well as those who provide AI tools for your use.

Construction projects and the AI used in them can vary, so the relevance of the points above will depend on the nature of the project, the AI tool and the risks arising. If the risk is low, it may be that the parties may decide to proceed with their normal contractual approach. But if AI input into the project is (or could be) very significant, the risks will be different and a more bespoke approach may be needed.

For now, there’s no standard, best-practice approach to address AI in construction contracts. With AI capabilities growing exponentially and with court cases taking years to progress, such standards may not appear for some time. As such, a common-sense approach is needed: consider the project and, rather than assume AI won’t be used (which seems unrealistic in the current climate), discuss the use of AI and understand the risks, then as appropriate, agree risk allocation and mitigations with the other party, and reflect that carefully in your contracts as needed.

Vicky McCombe, a partner at Womble Bond Dickinson who specialises in construction law, and Michelle Essen, a legal director in Womble Bond Dickinson’s construction and engineering team, also contributed to this article.