What do startups need from the Artificial Intelligence Act?

March 16, 2022
Artificial Intelligence Act

For many AI startups, the global challenges we are currently facing (COVID-19, climate change) are opportunities to use technology to improve people’s lives. At the same time, they are building technologies that are currently being regulated by EU policy makers. In this blog post we share key priorities for policy makers to consider in order to ensure the AI Act empowers startups. 

Presented almost a year ago, the Artificial Intelligence Act’s (AI Act) objective is to promote the uptake of AI in Europe by placing safe, trustworthy AI systems on the market.  We share the ambition of policymakers to create a harmonised Single Market for AI systems which will create a clear pathway for startups. EU policy makers can ensure that the AI Act will give startups the legal certainty they need to thrive by implementing the following principles:

Fast track procedure for sandboxes and ex-ante conformity  

Startups want to be able to enter markets quickly and easily, be it in low or high risk fields. Regulatory sandboxes are a great tool for startups to develop innovative products while ensuring the safety of their customers. To ensure harmonisation, sandboxes should be implemented in all Member States, with enough resources allocated to them to operate smoothly. Sandboxes should also be easily accessible by startups. In addition, conformity assessment procedures are often lengthy procedures that can result in the close of business for a small startup. It can take up to 6 months or a year to get a certification. Priority should be given to establishing fast track procedures, so startups can apply and utilise the sandboxes or have their AI system certified in less than four weeks. 

Proportionate high risk requirements 

Policy makers should incentivise highly ambitious, high risk startup entrepreneurs – these are the innovations that disrupt and transform our everyday lives.  The Artificial Intelligence Act can be an innovation catalyser by taking into account the reality of small, high potential companies. However, for now the Commission’s proposal does not also take into account the limited resources of small scale providers. Requirements regarding robustness, accuracy and cybersecurity should be flexible enough to adapt to any AI system in all sectors. Having data sets that are free of errors and complete is almost impossible to achieve in practice. It is therefore crucial that the AI Act reflects what AI entrepreneurs can realistically and practically comply with. In addition, ensuring proportionality will be key for allowing startups to develop innovative AI systems in high risks sectors. 

Clarity and precision for high risks operators

The AI Act has the potential to be a future-proof legislation. To do so, policy makers need to add more clarity and precision for high risk applications and their use. Innovation in high-risk field is not akin to a high-risk use. Policy makers should take into account the use case of the AI system when assessing if it should be categorised as low or high risk. Focusing only on the sector would disincentive startups to innovate in high-risk fields, such as healthcare or education, and would negatively impact the wide array of new services and technologies that AI can offer to EU citizens. 

Fast track procedures, proportionate requirements and legal clarity are the key ingredients that will allow startups to develop and offer their groundbreaking technologies to European citizens. The Artificial Intelligence Act rules will define the future of the development of artificial intelligence’s systems in the EU. Let’s make sure the legislation is implementable and understandable from day one by the most innovative actors of the EU economy – startups!