What exactly does the AI Act entail?
The AI Act aims at providing regulatory clarity and detailed measures for governing AI applications based on EU values, norms, and standards. It categorises AI systems based on their risk level to society, ranging from minimal risk to unacceptable risk and imposes stricter requirements for higher-risk categories, including transparency obligations, accuracy, and cybersecurity measures. The Act’s main goal is to ensure AI safety, transparency, and accountability within a human-centric ecosystem.
When will the AI Act come into effect and what does the phased implementation look like?
The Act officially entered into force on 1 Aug 2024. Based on the current estimated implementation phases, in a nutshell, after month 6 prohibition regarding unaccepted and prohibited AI systems will take effect; after month 12, general-purpose AI models (GPAI)-related rules and obligations will apply; after month 24, in 2026, AI Act aims to be fully applicable (see here for more information).
Why is the AI Act relevant for cities and municipalities?
Within the EU: Given the AI Act’s central position within the EU’s digital policy framework, the understanding of, and compliance with, the Act is compulsory for public sector entities – from the local to (inter-)national level. Cities and municipalities can be classified as deployers and/or providers of AI-enabled systems and thus must ensure compliance with the AI Act and linked regulations.
Beyond the EU: Although the AI Act does not currently apply in other jurisdictions, it may nonetheless inspire regulatory efforts and/or lead to potential (even partial) adoption in other parts of the world, especially in like-minded countries.
What are the main benefits that cities and municipalities can expect from the AI Act?
Although implementation challenges will exist in the foreseeable future, the Act is a step forward for providing regulatory clarity when developing, deploying, and/or managing AI-enabled systems. Cities and municipalities can benefit from the introduced structures and procedural roadmaps, addressing some of the key challenges surrounding the use of AI – from enhanced transparency fostering citizens’ trust to having clearer specifications for procurement.
What potential difficulties might cities and municipalities encounter when implementing the AI Act?
At the macro level, there exist two main challenges. First, the navigation of the existing ambiguity of the Act’s implementation and enforcement, especially as many implementation details of the Act are yet to be specified. Second, to achieve the necessary political and organisational understanding of what the Act entails and internal willingness and means to support implementation.
How can cities and municipalities prepare for and manage the implementation of the AI Act in their daily operations?
Based on the timeline presented above, the following three steps should help manage the arrival of the AI Act. First, prioritise building the institutional knowledge on what the AI Act entails and how it relates to your existing individual regulatory environment. Second, audit deployed systems making use of AI and (self-)classify their risk level. Third, keep up with the national level developments, in particular through your Market Surveillance and Notifying Authorities and how they can support you. With AI becoming a ubiquitous component of your digital ecosystem, also seek support from peers and stakeholders as pointed out below.
What support and resources are available for cities and municipalities to help them comply with the AI Act?
The establishment and/or appointment of institutional support on EU and national levels is still in an early stage. It is important to proactively monitor progress to seek further clarification and support, especially from the national level. Cities and municipalities can also actively participate in workshops/discussions and networks/communities such as OASC to learn from – and with – your peers and stakeholders. In addition, active engagement with broader initiatives such as the EU AI Pact and other EU-funded projects on this topic may provide valuable support and opportunities.
Are there best practices or successful examples of cities that have already begun implementing the AI Act?
With more details on the implementation of the AI Act yet to be specified, it is too early to discuss the best successful cases of city implementation. However, there are several early-stage developments at the local level that aim to promote a more transparent, trustworthy, ethical, and human-centric approach to AI, such as Amsterdam, Barcelona, Berlin and Helsinki. Whether these existing local activities fully comply with the AI Act remains to be seen. However, they show that awareness of the issues exists on a municipal level and that some of the digital forerunners will likely help to navigate the implications of the AI Act for cities and municipalities going forward.
Our experts also want to remind cities and municipalities that they are not alone in this process. Irrespective of the complexity of one’s existing AI systems and/or future AI strategy, it helps to engage with a diverse set of peers to make sense of both the challenges and opportunities the AI Act holds within and beyond the EU.
For further information, please contact:
Ivy Yang (ivy@oascities.org)
Gert Hilgers (gert.hilgers@oascities.org)