The EU Artificial Intelligence Act: premature or precocious regulation?
The European Union’s Artificial Intelligence Act is essentially a product safety regulation. Risks can be assessed reasonably well for older-generation single-purpose artificial intelligence (AI) models. It becomes more problematic for the latest Generative AI models, like ChatGPT, that can be moulded for an almost infinite range of purposes. The AI Act tries to work around this with a general obligation to avoid harm for human users. General purpose AI (GPAI) models are a source of concern, especially when they become very large and pose systemic risks. GPAI developers must build in ‘guardrails’ that avoid harmful responses. However, there are many ways to circumvent these guardrails.
There is also vigorous competition between AI start-ups and big tech firms and no sign of emerging monopolistic gatekeepers yet. However, smaller firms need access to large AI cloud computing infrastructure where big tech firms dominate. Big tech firms can put AI models to work in their established business services to generate revenue. That leads to close collaboration agreements between start-ups and big firms that may tend towards integration. AI startups need to set up their own business services, often with riskier open-source AI models that become platforms where deployers can plug in their own applications. New AI-driven ecosystems are emerging that combine AI models with existing online services. The borderline between developers, deployers and users becomes a bit foggy.
The Act as it stands today is just the start of a long regulatory process. It delegates responsibility to the European Commission and its newly created AI Office to draft implementation acts and guidelines to address these challenges. These will drive enforcement of the Act and determine to what extent it will be a precocious instrument to stimulate trustworthy AI innovation or a premature innovation-smothering regulation.
Read more about the European Union AI Act in Bertin Martens' Analysis 'The European Union AI Act: premature or precocious regulation?'
The Why Axis is a weekly newsletter distributed by Bruegel, bringing you the latest research on European economic policy.
About the authors
Related content
Why should EU copyright protection be reduced to realise the innovation benefits of Generative AI?
Economic arguments in favour of reducing copyright protection for generative AI inputs and outputs
The licensing of training inputs slows down economic growth compared to what it could be with competitive and high-quality GenAI
The European Union AI Act: premature or precocious regulation?
As it stands, it is unknown whether the Act will stimulate responsible AI use or smother innovation.
Are new EU data market regulations coherent and efficient?
Technical restrictions on access to and re-use of data may result in failures in data markets and data-driven services markets.