AI in the sandbox

No one is likely to have missed the recent news reports and writings on AI, which has been progressively 'booming'. Europe, through its work on a new regulation on AI, is paving the roads while driving at the same time. So far, it is not easy to know what you can and cannot do when it comes to the use of AI.

Among other things, the AI Regulation will introduce requirements that AI systems must fulfil, which is welcome, because who wants a world full of AI that is unpredictable and unreliable? However, requirements can often make the journey to a finished product much more difficult and costly than it would be otherwise. It can then be tempting to put blinders on and instead try to run to the finish line as fast as all the other innovators, to gain ground and enjoy profits. Some may be lucky and get it all right, but luck does not build stable skyscrapers that people want to live in. The vast majority who rely on luck crash and end up with a pile of rubble and sky-high costs caused by damage.

The AI Regulation will impose requirements on security, transparency, data quality and risk management. There will be many requirements for those AI systems that are identified as posing a high risk when used. These could be systems in areas such as education, law enforcement, justice, or critical infrastructure. One tip is to look at the proposed AI Regulation now to get an idea of what to expect. Even if it is only a proposal, many benefits can be gained by staying up to date.

As with GDPR compliance, stakeholders will need to know the "how", "what" and "why" when it comes to AI systems. No one can pretend to be blissfully ignorant here. Those actors who have taken GDPR seriously and worked systematically with data protection will have a lot to gain from already established ways of working. Being aware of the basic principles of the GDPR automatically means "keeping the house clean". Having robust risk management processes in place today will pave the way for the requirements of the AI Regulation. Those with accurate and up-to-date data will benefit here too, as an AI system will never be better than the data it has been trained on. Accurate data gives a higher chance of accurate decisions and predictions than an AI system trained on old, irrelevant or even inaccurate data. Really good data will be a goldmine, now and in the future.

It may seem scary to sink your teeth into AI and let innovation flow when there are so many rules about what you can and cannot do. However, the AI Regulation also takes steps to promote innovation, as the proposal includes provisions on so-called regulatory sandboxes. These aim to bring together innovators and authorities in a controlled environment to promote collaboration and facilitate compliance with the forthcoming AI Regulation. The regulatory sandboxes will allow for additional processing of personal data for the development of certain AI systems in the public interest. For those who are qualified to access the sandboxes, this opportunity can provide access to incredibly good data to train their AI system on.

To those who are either AI developers, distributors, or users of AI systems: Don't wait for the AI Regulation to come into force! Don't make the same mistake as with GDPR, where so many started their compliance work too late. There is ground to be gained, trust to be created and, above all, incredibly groundbreaking AI systems to be developed or used for those who start in time. Get the right conditions now, create valuable data by embracing GDPR and start assessing risk. Shovels, buckets, and sandboxes are already available. The best sandcastle wins!

We at Fondia have a large team with high expertise in AI. Do not hesitate to contact us if you need help with your AI systems and all the regulatory requirements that come with it. You can find more information about our AI services here.