Trust by design: Why ethical AI will define the future of AEC tech

Earlier this week, I had the privilege of attending the Human–Robot Collaboration in Construction Symposium, with participants joining in person from across the world. The conversations were thoughtful, ambitious and at times genuinely inspiring.

What surprised me most was not the technology itself but the structure of the conversation. Governance and policy came last, almost as an afterthought.

That moment stayed with me. Not because the discussion lacked insight – it didn’t – but because it reflected a broader pattern across our industry. We are enthusiastic about innovation, eager to test new tools and rightly focused on productivity, performance and budget. Yet governance, ethics and accountability are still too often treated as secondary concerns. In an industry where safety, public trust and long-term responsibility are foundational, that imbalance matters.

Over the past few years, artificial intelligence has moved quickly in AEC, not just in capability, but in where it shows up. It is no longer confined to research programmes or innovation showcases. It now sits inside everyday tools used for early-stage design, programme forecasting, site monitoring and asset operations. Often, it isn’t even labelled as “AI” anymore. It’s simply presented as a smarter feature, a faster workflow, a better recommendation.

And that is exactly why ethical AI matters now.

In my experience, in many industries a bad AI recommendation is merely an inconvenience. In construction and infrastructure, however, it can affect safety, claims, compliance and service continuity. When a system suggests a programme change, flags a hazard or predicts equipment failure, the immediate question is simple: who is accountable if it’s wrong or if it’s followed without challenge?

AEC relies on professional judgement, standards and clear responsibility. Black-box systems that can’t be explained or contested sit awkwardly in that culture. Trust isn’t about believing the model “works”; it’s about being able to justify decisions to clients, regulators, insurers and the public.

Three practical issues come up repeatedly.

Transparency: Many models, especially deep learning, are hard to interrogate. They can be accurate, but not necessarily explainable in a way that stands up to scrutiny. In a contract-heavy environment, “the model said so” is not a defensible position.

Bias: AI learns from historical data and lets face it – our construction data is messy. Schedules can be optimistic, incidents may be underreported and quality records can be inconsistent. If a model absorbs those patterns, it can reproduce them at scale, while producing outputs that look confidently precise.

Privacy: Site imagery, sensors and behavioural data can quickly drift from safety improvement into surveillance. Clear rules on collection, access, retention and purpose are essential if you want workforce and client trust.

Ethical AI, in this context, isn’t abstract. It’s about protecting workers, safeguarding clients and maintaining public confidence in the industry.

This is where the EU AI Act stops being a headline and becomes a delivery issue. It takes a risk-based approach: the higher the potential impact on health, safety or fundamental rights, the stronger the obligations.

For AEC, that matters because certain uses, depending on context, may be treated as high risk. AI used for safety management, workforce decisions or essential infrastructure operations is exactly the kind of application regulators want governed properly. The implication is straightforward – organisations must be able to demonstrate risk controls, data quality, documentation, cybersecurity and meaningful human oversight.

The Act doesn’t ban AI in construction. It changes the burden of proof. You can innovate but you must be able to show that what you built (or bought) is responsible.

Policy isn’t only regulation; it’s also investment. Through Horizon Europe, the EU is investing over €1 billion a year in AI from 2021 to 2027, with an emphasis on trustworthy, human-centric development. Programmes such as GenAI4EU with funding calls planned for 2025 and 2026 are intended to push generative AI into real industrial use, not just demonstrations. In parallel, AI-in-science initiatives like RAISE are strengthening the foundations: methods, models and compute, that will spill into engineering, simulation, and infrastructure planning.

For AEC, the takeaway is that public funding helps set expectations early. It nudges the market toward transparency, robustness and accountability, often before those requirements show up in procurement language.

The good news is that “ethical AI” can be operational. Frameworks such as the OECD AI Principles and the NIST AI Risk Management Framework translate values into steps most project teams already recognise – define purpose, understand limitations, build oversight in and monitor performance over time.

In my opinion, security requires the same discipline. As AI becomes embedded in operational systems, risks like data poisoning, model drift and unintended inference can’t be treated as edge cases. ISO/IEC 27001 remains a baseline, while ISO/IEC 23894 and ISO/IEC 42001 provide AI-specific guidance for repeatable controls, so trust is built into delivery, not added after the fact.

Looking back at the symposium, I was encouraged by how far the technology has come. But trust isn’t an automatic by-product of innovation. It has to be designed, governed and maintained.

The future of AI in AEC won’t be defined by who deploys the most advanced model. It will be defined by who can demonstrate ethical, secure, accountable systems, tools that support professional judgement rather than obscure it. In our industry, trust by design is what will determine whether AI becomes a lasting foundation or just another passing experiment.

Lilian is a visionary architect turned Digital, BIM, and AI Lead, with over 25 years of transformative experience across both public and private sectors. Renowned for her expertise in Digital Construction and BIM Management, she is a recognised industry expert with a diverse portfolio that includes projects in aviation, education, commercial, defence facilities and spaceports.

An industry leader in data-driven design, BIM workflows and digital transformation, Lilian excels in developing innovative digital strategies and solutions for complex projects. She leverages advanced technology, automation and artificial intelligence to drive efficiency and innovation in the construction industry.

In addition to her professional accomplishments, Lilian is a dedicated guest professor and mentor, passionately advocating for digitalization within the AEC industry. Her commitment to education and mentorship is shaping the future of digital construction on a global scale.