AI set to drive construction in Hong Kong SAR

In early February, a group of senior executives and industry leaders from across the public and private sectors gathered at a roundtable held in Pinsent Masons’ Hong Kong office to share insights on both the use and future of AI in Hong Kong’s construction industry and its wider built environment sector.

The Hong Kong Special Administrative Region (SAR) does not currently have a dedicated overarching artificial intelligence regulation. Instead, various government bodies have implemented or issued sector/ context-specific initiatives, guidelines and policies, and generally promoted the use of AI by the built environment sector. There are numerous examples of the private corporates adopting AI within the built environment to improve the efficiency and safety of construction as well as management of the as-built asset.

Technology is already at the heart of the government’s ‘Construction 2.0’ policy, which aims to strengthen and maintain Hong Kong SAR's regional leadership position in construction.

As public sector investment in Hong Kong SAR continues to grow, it is imperative for the construction industry to align with the government’s evolving priorities. The roll-out of AI-powered platforms to harvest data as well as promotion of AI tools – such as BIM, safety equipment and robotics – offers countless opportunities to enhance productivity and drive efficiencies across the industry.

As participants at the roundtable recognised, it is now incumbent on industry itself to embrace digital tools to address safety, efficiency and skilled labour shortages. As in other jurisdictions, the successful adoption of AI in Hong Kong’s construction industry will require significant levels of investment, training and manpower to ensure stakeholders can collaborate effectively to drive forward an integrated, data-driven approach.

Participants expressed concerns regarding data security, potential data leakage and confidentiality issues arising from the use of AI, as well as known difficulties in harvesting data and the reliability and accuracy of AI outputs.

Making AI mandatory

Recent government initiatives to make AI use a contractual requirement of public projects are a critical first step in the right direction.

The Development Bureau (DEVB), the department that oversees Hong Kong’s construction industry, has been at the forefront of implementing ‘Construction 2.0’. One of the DEVB’s flagship policies has been mandating capital works projects to use a centralised, web-based platform called the Digital Works Supervision System (DWSS) to collate construction works information and manage workflows of site activities and contract management.

Since 1 April, public capital works projects with an estimated budget exceeding HK$30 million (approx. US$3.82m) have been required to use the DWSS for the design and construction phase. Although this is a contractual requirement rather than a statutory regulation, the policy makes it mandatory for contractors, sub-contractors and consultants to use the DWSS when participating in these types of projects.

The DEVB is also developing the Integrated Capital Works Platform (iCWP) to collect and consolidate data from all DWSS and other digital systems in capital works contracts for continuous monitoring and data analysis.

These changes herald a significant shift for Hong Kong’s construction industry, which is dominated by public sector projects. The government is responsible for building all the region’s major infrastructure projects, from railways and roads, hospitals, schools, community facilities and public housing.

These steps to embrace digitalisation are also part of broader initiatives by the government to promote the adoption of AI technologies across the construction industry. Since 1 April, public works projects over HK$30 million have also been required to use AI-driven drones – Highly‑Effective Construction Robots (HECR) – from DEVB’s inventory for site inspection and monitoring.

Data troubleshooting

These measures reaffirm the government’s commitment to ensuring AI serves an increasingly integral role in assisting construction management and driving the workflow integration and business processes of many of Hong Kong’s largest and most high-profile construction projects over the coming decades.

However, roundtable participants pointed to a number of challenges surrounding the security, confidentiality, reliability and shareability of data in Hong Kong’s current built environment context.

AI models need machine readable digital data in consistent data platforms, but widespread adoption across the industry is hampered by the fact that there is currently no central database of datasets available for private companies to use to train AI platforms. Although this is likely to change, it is unclear what level of data will be shared publicly.

Many organisations have accumulated significant project data over the years, for their own use. Participants noted that there remained gaps in usable data from their existing records or they are still working on how to identify which data points are relevant and should be prioritised for training AI models to support project management, scheduling, and cost management.

Concerns were also expressed about how AI models utilise proprietary data and whether such data may be shared with external parties. Organisations remain cautious about cloud‑based training environments and generally favour private‑cloud or on‑premises solutions.

Where AI software is trained using datasets obtained from third parties, there remain concerns about the accuracy and reliability of its outputs. This is exacerbated by the so-called “black‑box” nature of AI, which can make it difficult for stakeholders to understand how conclusions are reached or to confirm their veracity.

Currently, most AI firms targeting the construction industry are start‑ups that lack sufficient datasets and rely on private companies to provide data for training; raising concerns about data usage, further exploitation, and the risk of data leakage. For many private companies, there is often no commercial incentive to undertake extensive AI testing and piloting, or even to share training datasets with other organisations. Such data is considered confidential and sensitive. The risk of potential data leakage or breaching of proprietary data also puts them off doing so from a legal risk and even ‘trade secret’ perspective.

Time is also seldom on their side. The project lifecycle for a design consultancy project may allow only around six months for AI testing, which may be insufficient for meaningful AI training. Even where piloting is possible, it may take several years to properly train an AI model, which can be prohibitively lengthy for start‑ups requiring substantial upfront capital. Many organisations at present do not have internal or external personnel with sufficient knowledge of data collection to manage such processes effectively.

What’s next for data sharing

Because of these concerns, data gathered in construction projects by individual companies and statutory bodies has not been shared between them. This in turn has inhibited AI models from being trained effectively, thus limiting the improvement of accuracy and reliability of AI models.

As the largest client in the construction industry, the government holds extensive project data and is well placed to collate large project-related data sets. If such data were systematically collected, anonymised and made available to the industry for AI training – as in the case of data collected through the iCWP – this could have significant benefits for both the construction industry and AI developers.

The roundtable participants considered that utilising AI models trained on publicly available datasets, alongside more specialised, domain‑specific AI models trained on proprietary corporate datasets, could be the best path forward for Hong Kong SAR.

This would allow AI companies to use government datasets for training, while still enabling private companies to adopt AI products trained on such data, or co‑develop bespoke AI solutions further trained using proprietary datasets. This approach would improve accuracy while reducing the risk of producing hallucinations or inadvertent cross‑sharing of private datasets.

Data-sharing agreements could also be used to help organisations manage their data more effectively, while retaining their data sets’ commercial value, confidentiality, privacy, and security.

To facilitate this, organisations will have to invest in data consolidation, harmonisation, metadata management and accessibility. Both public and private entities should ensure that their data is properly structured and ready for AI to harvest, analyse and learn from if such an approach is to be applied effectively across the construction industry.

Regulatory framework and governance

Hong Kong SAR leans towards a context-based ‘soft law’ approach primarily driven by its Digital Policy Office (DPO), the Office of the Privacy Commissioner for Personal Data (PCPD), the Commerce and Economic Development Bureau (CEDB) and the Intellectual Property (IP) department. In this way, its approach differs from that taken by other jurisdictions – for example, the European Union, which took the route of a ‘hard law’ approach when enacting the EU AI Act back in 2024.

In July 2024, the CEDB and the IP department undertook a two-month public consultation to address the need to update copyright laws in response to advancements in AI technology. The results are still awaited.

In July 2024, the DPO published the Ethical Artificial Intelligence Framework (Version 1.4). Initially intended for public bureaux and departments, the framework is now also available for the private sector. As voluntary guidance rather than mandatory rules, the framework encourages organisations to embed ethical principles, assess AI risks, and integrate AI governance into existing risk management and project governance processes.

In March 2025, the PCPD also published a checklist on guidelines for the Use of Generative AI by Employees, which encourages organisations to implement internal policies and monitoring mechanisms to uphold principles such as defining the appropriate scope of AI use, protecting personal data privacy, preventing biases, and ensuring compliance with legal and ethical standards, thereby promoting responsible AI practices within the workplace.

In April 2025, the DPO also released the voluntary Generative Artificial Intelligence Technical and Application Guideline, which aims to provide practical guidance for technology developers, service providers and users of AI. It covers the scope and limitations of applications, potential risks and governance principles of generative AI technology, including technical risks such as data leakage, model bias and errors that need to be addressed.

Currently, sketches, drawings and plans, are covered by copyright under the existing Copyright Ordinance. However, if they are wholly AI-produced, then that can potentially leave contractors and developers without IP rights to the output, which can be problematic from a legal perspective.  

In a 2024 consultation (61-page / 1.7MB PDF), the administration explored enhancing the Copyright Ordinance to ensure adequate protection for AI-generated works and the implications of AI on existing copyright frameworks. It examined issues such as liability for copyright infringement involving AI, the potential introduction of specific copyright exceptions text and data mining and computational data analysis for AI training.

Robust data governance frameworks will be critical to prevent sensitive datasets from being disclosed to third parties and to mitigate potential liability risks. Currently CI operators – which operate critical infrastructure in Hong Kong SAR – have different statutory obligations to AI service providers when it comes to conducting risk assessments and audits. However, the roundtable discussions made it clear that it incumbent on all industry stakeholders to establish adequate safeguards to monitor AI performance, as well as guardrails to guarantee data security regardless of the project or development in which they are engaged.   

While the regulatory environment for AI is much more advanced in Singapore and mainland China, all of these and other initiatives being taken by the administration in Hong Kong signal its proactive commitment to AI adoption in the built environment space.

Mainland China’s rapid industrialised construction is already influencing infrastructure developments in Hong Kong SAR, as evidenced by the rapid rise of modular construction in the region. Amid recent reports of Chinese companies integrating AI across construction machinery, manufacturing, management and robotics, it seems likely that the future of construction in Hong Kong SAR will be increasingly underpinned by AI as well.

Future outlook

The roundtable made the overall direction of travel abundantly clear: AI is expected to play an increasingly significant role in construction in Hong Kong’ SAR. Targeted public sector piloting and funding mechanisms are expected to accelerate AI adoption across the industry. Industry players need to be poised to seize these opportunities quickly, while mitigating and managing the risks that AI may bring.

AI has the potential to influence organisational design, workflow integration and business processes, and is expected to be widely adopted across the construction industry in the near future. Organisations should develop AI literacy through training programmes, cross‑functional teams and recruitment of specialists in this field.

In a blog post published in early April, the chairman of Hong Kong’s Construction Industry Council (CIC) proclaimed that 2026 would be the ‘Year of AI’ for Hong Kong’s construction industry and this would require “empowering” the entire workforce – from workers, engineers to managers – to use AI-powered tools to “work more safely and efficiently.” The message was clear: there must be sufficient “buy-in” from all industry stakeholders to make this strategy work.

While AI adoption remains uneven, the latest measures will increase its momentum in public sector works and may also cause contractors and other stakeholders in the private sector to recognise its value for other large-scale projects too.

What is clear is that organisations that engage early are likely to be best positioned to shape and benefit from these developments. Those that view AI holistically – as part of an operating model rather than a standalone technology – are also likely to derive the greatest long‑term value.

Co-written by Dr. Nils Rauer and Kingston Yeung of Pinsent Masons.