The Planning and Regeneration Committee met to discuss the widespread use of Artificial Intelligence (AI) in London's planning system. Experts highlighted that while AI adoption is spreading, it remains slow due to various barriers, with a significant divide between the public and private sectors in its utilisation. Key challenges identified include data accessibility, the need for clearer regulatory frameworks, and the potential for AI to exacerbate existing inequalities if not implemented thoughtfully.

The committee heard from a range of experts about the current state of AI in planning. Professor Lauren Andres from UCL Bartlett School of Planning noted that while general-purpose AI tools like ChatGPT are widely used, planning-specific AI tools are still in the early stages of adoption, often limited to small-scale pilot projects funded by initiatives like the MHCLG PropTech Innovation Fund. Nissa Shahid of Arup observed that while exposure to AI is high, its use is often confined to basic tasks like summarising documents, rather than leveraging its full potential for complex problem-solving. Professor Samer Bagaeen highlighted a marketplace of AI applications, with the public sector lagging behind the private sector in adoption due to risk aversion, budget constraints, and a lack of visionary leadership.

Significant barriers to AI adoption in the public sector were discussed, including the reliance on non-machine-readable documents rather than structured data, the high cost of acquiring quality data, and a lack of funding for experimentation. Professor Mike Raco identified legal and regulatory hurdles, such as the inability of AI to provide 100% accuracy and the challenge of assigning responsibility when AI makes errors, leading to fears of judicial review. Governance fragmentation across London's 33 boroughs and a verification measurability gap where human planners struggle to verify the vast amount of AI-generated information were also cited as major obstacles. Furthermore, the insurance industry's reluctance to recognise AI as a source of evidence presents a significant challenge for its widespread use in planning.

The discussion also touched upon the ethical considerations of AI in planning, including data ownership and the potential for bias in algorithms. Professor Bagaeen stressed the importance of LPAs retaining control over their data, warning against a data race where external suppliers could exploit data held by local authorities. The ethical implications of automated decision-making were also raised, with concerns about legal gaps and the need for clear guidance, particularly in relation to the Local Government Act 1972.

The committee explored the potential benefits of AI for developers, including cost reduction and faster application processes. However, experts cautioned against focusing solely on speed and cost, emphasising the need to maintain creativity and ensure AI complements, rather than replaces, human judgement. The potential for AI to democratise the planning process by enabling citizens to better understand and engage with applications was also discussed, alongside the risk of agentic flooding where AI could be used to generate large volumes of responses, potentially overwhelming planning departments.

Recommendations were made for the GLA to take a leading role in developing a clear vision and strategy for AI in planning, including establishing shared data standards and infrastructure across boroughs. The need for comprehensive training for planners on AI tools and risks was highlighted, alongside the development of robust acceptable use policies. The potential for AI to transform planning into a more dynamic, data-driven process, moving away from periodic reviews towards continuously updated evidence bases, was also explored. The committee concluded by acknowledging the rapid pace of AI development and the need for ongoing adaptation and regulation to harness its benefits while mitigating its risks.