Artificial Intelligence (AI) needs to be regulated, Tesla and SpaceX CEO Elon Musk said in 2020. Musk proposed that individual governments as well as global organisations like the UN should come together to regulate the development of advanced artificial intelligence.
Soon after his tweet, the European Union proposed a plan to regulate “high risk” AI systems. Now, the European Commission has revealed its plans to build a regulatory sandbox to foster innovation in the field of AI. The framework proposed by the commission could pave the way for safe and regulated AI development.
What is an AI regulatory sandbox?
The artificial intelligence act envisaged by the European Commission recommends setting up coordinated AI ‘regulatory sandboxes’ to foster innovation in AI across the EU. The proposal recommends the regulatory sandbox will act as a tool allowing “businesses to explore and experiment with new and innovative products, services or businesses under a regulator’s supervision.”
The central idea is to create a controlled environment for testing innovation in the field of AI. This controlled environment will also make it easier for regulators to better understand the technology. Basis discussion with academics and stakeholders, the European Parliament has called for introduction of regulatory sandbox instruments in several resolutions.
“Academics and stakeholders have commented on the proposal, touching, in particular, on issues regarding the lack of liability protection for sandbox participants, the need for a more harmonised approach to AI regulatory sandboxes, and the interplay between AI sandbox and EU data protection rules,” the European Parliament explains in its briefing.
What will be the role of AI regulatory sandbox?
There is no agreed definition for regulatory sandboxes but the European Parliament is referring to this being regulatory tools that allows businesses to test and experiment with products, services or businesses. However, it will be done under the supervision of a regulator and for a limited period of time. The sandbox will play following role:
- Foster business learning: The primary role of AI regulatory sandbox will be to support the development and testing of innovations in a real-world environment.
- Support regulatory learning: The regulatory sandboxes are also envisaged to act as formulation of experimental legal regimes to guide and support businesses in their innovation activities under “the supervision of a regulatory authority.”
The EU is no stranger to the sandbox approach and such sandboxes operated by financial regulatory authorities are now widely used in financial technologies. A World Bank study shows that more than 50 countries are currently experimenting with fintech sandboxes. The sandbox approach has expanded to transport, energy, and health as a testbed for new innovation.
What will be the benefits of AI regulatory sandbox?
- Regulatory benefit: The biggest benefit will come in the form of regulators better understanding innovative products built by AI-fuelled companies. It will also allow regulators to develop “adequate rule-making, supervision and enforcement policies.”
- Mitigation of risk: For innovators, the benefit will be to test their new products in a controlled environment and mitigate the risks associated with them. They will also be able to reduce the time to market for their products with this sandbox approach proposed by the European Parliament.
- Safer products: For consumers, the benefit will come in the form of new and potentially safer products. The European Parliament wants to foster innovation and consumer choice while also ensuring innovation happens in a safe way.
What are the concerns with regulatory sandboxes?
- Misuse: The primary concern is around misuse of regulatory sandboxes, forcing regulators to lower safeguards and requirements to attract innovators.
- Innovation over safety: There is also a growing concern among critics that this approach will prioritise innovation over putting adequate safeguards to protect the public and consumers.
- Slowdown: There is also concern among private actors that the sandbox approach will interfere or slow down genuine innovation.
- Fragmentation: The regulatory sandbox also runs the risk of further fragmentation of the EU single market if the parameters differ significantly in member states.
What else do you need to know about the draft AI act?
Apart from recommendation for a regulatory sandbox, its benefits and concerns, the draft AI act also envisages setting up coordinated AI sandboxes at the national level. Under its governance and supervision, the European Parliament is recommending establishment of common rules to ensure uniform implementation of the sandboxes across the EU.
The proposal also makes it clear that participation in a sandbox experiment will not result in exemption from liability for participants. “Participants in AI regulatory sandboxes would therefore remain liable under applicable EU and Member State legislation for any harm inflicted on third parties as a result of the experimentation taking place in the sandbox.”
The draft AI act also clarifies the interplay between the new horizontal rules for AI and the applicable data protection rules. It requires the member states to associate the national data protection authorities or other competent authorities with the operation of AI regulatory sandboxes.