03 July 2026
The demon is energetic and really is off the bottle.
RIBA's AI Report 2025 (released June 27, 2025) found that over half of British architect practices (59%) actively use the Geneai (Genai) tools. This is an almost 50% increase in previous surveys from 2024.
The new consensus says the profession will be strengthened by AI. This is considered a technology that can increase productivity and creativity, but there are also concerns about the impact on future employment outlook and fees.
However, increased usage increases the risk.
May Winfield, the global director of commercial, legal and digital risks at Buro Happold and a leading expert on the legal aspects of digital risk management, says the power of AI will not blind these risks.
She warns that potential legal and liability risks may lie under the radar and that practices implementing digitalization as well as AI may not pay sufficient attention.
https://www.youtube.com/watch?v=b43mt15rs9y
Do architects lose copyright to data or design when using AI?
Entering practice data into a publicly generated AI tool is so that you can say “to enter it in a publicly available box or forum.” Depending on how the tool works and how the data is stored, you may be able to extract the architect's data, images, or designs using the appropriate prompts.
The architect clearly lives on risk. 69% of respondents to RIVA reports believe that AI increases the risk of being mimicked. Therefore, it is important to understand how tools store and utilize someone's data. Are models trained with architect data? Has the architect lost control or copyright of the data or design when entering into a generator AI tool or AI agent?
“Answers can depend on terms of use, how the model is trained, and other factors, so it is important to review terms of use and have informed discussions with internal IT teams, legal teams, specialized advisors, and AI providers. “This allows for an informed decision, for example, whether some data should not be entered into the tool to maintain copyright.”
She continues. “We've all seen headlines in lawsuits against AI companies over copyright violations. Recent headlines show some positive victories in their favor. The most notable case shows that the principle of fair use for “fair use” is much broader than the judicial powers of the UK. How other corresponding AI regulations affect and apply the rights of the parties. ”
However, we believe that this watch is a matter of the legal rights and copyright use positions where it is not expressly permitted/unauthorized by the author/owner.
Several large tech companies provide compensation to AI users for potential copyright infringement. We assume that each company has its own language and it will be interesting to see how their terms are actually applied and interpreted.
What are the issues surrounding confidentiality and client or contract restrictions?
Regarding confidentiality, May states that it is important to discuss with legal advisors and digital specialists due to client requirements, terms of contracts, or other restrictions.
The language of the terms of use of AI tools should also be considered to understand how confidentiality is maintained and whether this is acceptable to architects.
“It is worth noting that references to AI in bidding documents. This extends from prohibiting the use of AI to prepare AI or other requirements within the bidding/contract scope, to ensure that the appropriate experts in the organization provide input related to bidding returns,” May suggests.
What is ai 'Hallucinations' and how can architects protect them?
AI hallucinations refer to cases where the AI model produces wrong, meaningless, or simply manufactured output. However, hallucinations are known phenomena within genai tools, and they may appear completely plausible.
It may suggest that organizations can partially mitigate such risks by constantly putting humans in loops for checks or by limiting the use of generated AI tools to places where hallucinations do not produce significant consequences.
Organizations should also discuss their intended use of AI tools, whether it is part of the scope of work/services or deliverables with insurance brokers, to ensure that they are fully insured if there is a claim related to the use of AI tools.
Also relevant to this discussion is how the architect covers, for example, if data is lost or compromised when using Genai tools, this causes problems with the project.
According to May, it's always better to talk to an insurance broker to make it clear than to discover that there may be potential issues with insurance coverage.
She also points out that internal guidance, policies and training are important to ensure awareness and best practices in the organization.
Do AI models have ethics and bias security and issues?
All practice is ideally comfortable, and you need to be clear about how data entry is stored (or not stored) and where it is located for security purposes, says Mays. It is often helpful to discuss with your Genai provider, internal IT team, and digital specialists in your practice to understand how security steps are enforced and whether training or guidance should be implemented in practice to ensure excellent practice compliance.
Some commentators have expressed concern about biases that may be present in the historical data that the model is being trained, thus affecting the response to AI tools prompts. Again, even if you put humans in a loop to check for such biases and inappropriate historical data, there are often the best ways to manage such problems.
We also recommend that you check that your organization's AI usage complies with general data protection regulations (GDPR))).
Thank you to Maywinfield, global director of commercial, legal and digital risk. This article is not legal or professional advice and readers should obtain professional advice before acting on its content.
Text by Neil Morris. This is a professional feature edited by the RIBA Practice Team. Send us your feedback and ideas.
RIBA Core Curriculum Topics: Design, Construction, Technology.
As part of the flexible RIBA CPD program, specialist functions are counted as microlearning. Details about the updated RIBA CPD core curriculum and how you fulfill CPD requirements as a RIBA charter member.
