The year 2023 was marked by, among other things, the approval of the world's first comprehensive AI regulation by the European Union. Although, as our colleagues at MCPRO warn, it still needs to be ratified by the European Parliament and the EU Council and is not expected to be in force until 2026.

The so-called AI Act was first introduced more than two years ago, so it is an unprecedented step forward for developers of AI systems, such as ChatGPT, to start meeting basic transparency requirements (unless they were free and open source). The regulation has focused on key aspects such as biometric surveillance and the characteristics of functional models.

But how have Spanish startups that work with or are based on Artificial Intelligence taken this draft? We wanted to talk to some of them to get their professional views on which aspects they like more and which ones they like less, as well as what they miss in this first regulation.

Sergio Álvaro Povedano, Country Manager of Contents.com in Spain and Portugal:

"The big problem with the law is that it is not being implemented globally".

For me the most attractive part of the law is the drastic support that Generative AI has received, but "with vigilance", for us this is something that was fundamental and that since our beginnings we have taken into account in all our developments. The best part of this law lies in the prohibitions on the use of this technology in certain fields, as for some uses of AI, the risk is considered unacceptable because of its potential threat to citizens' rights and democracy.

Artificial intelligence (AI), once only available to experts, has started to become commonplace among the people thanks to developments such as Open AI, Google's Gemini and WhatsApp's upcoming AI. This increased accessibility, however, has brought with it concerns about the potential dangers of AI.

Influential figures such as Elon Musk and Bill Gates have expressed concerns in this regard, especially regarding the mishandling of this sophisticated technology. AI can offer significant benefits, but it can also be a threat if used inappropriately.

On the other hand, the worst law lies in three fundamental aspects:

In terms of entrepreneurship, new developments and implementation of this technology; it will be a huge brake, and we will even discard projects before inventing them, something that will clash head-on with the evolution of this technology. In fact, the law to regulate AI in Europe is already generating discrepancies within the sector, as several actors in the sector point out that it could slow down innovation and the competitiveness of European companies, as the high commissioners of important countries such as France and Italy have commented.

It leaves the door open for certain sectors to use biometric surveillance systems, such as the public security sector or national security, which in my view is a serious infringement of the right to privacy. Of course it is important, but where is the limit, and this is what brings me to my third point.

The big problem with the law on AI regulation is that it is not being carried out on a global level, a unified law across all countries and their different actors, which can mean a huge brain drain between countries that exercise a more stringent law versus others that are less stringent.

Do you miss anything in this regulation Sergio Álvaro?

What I miss most is real regulation that combats bias in a much more concrete way. Generative AI models that focus on text generation, such as GPT, are called large language models (LLM). They have been trained on a large number of texts, including those from various epochs, even centuries ago. This training data may contain works by influential figures who advocated slavery or who held racist views.

This therefore implies that different kinds of biases can occur, so this issue should be much more present in the new law. Addressing these and other types of discrimination in training datasets is crucial to developing fair and accurate AI models that represent the perspectives of different populations.

Contents.com is a generative AI platform that helps create text, audio, video and animation, ensuring the highest standard of quality, allowing it to make a difference in the field of AI platforms.