Share This Article
Last week marked a pivotal moment in the history of digitalization as the EU Parliament agreed to the first legislation on artificial intelligence in the world, the AI Act.
The new legislation aims to regulate AI according to its potential for harm and includes stricter rules for foundation models such as ChatGPT to prevent intentional manipulation of content and the use of emotion recognition software in certain sectors. The EU AI Act outlines general principles for all artificial intelligence models, including the need for human supervision, technical robustness and security, data privacy and governance, transparency, social and environmental well-being, diversity, non-discrimination, and equity.
However, if the stringent EU approach outlined in the version of the AI Act that has been agreed upon is not followed by other jurisdictions like the US and China, the Act might slow down the pace of innovation within the European Union.
Furthermore, the provision adopted to address AI systems’ potential use of copyright-protected material has raised some concerns. Companies deploying generative AI tools must disclose any copyrighted material used to develop their systems, but a clearer provision on the matter would have been beneficial. For instance, the EU Copyright Directive already provides the text and data mining (TDM) exception, and a cross-reference to that in the AI Act would have cleared out any possible misinterpretation of the scope of the exception.
Additionally, the current version of the AI Act does not address privacy issues raised by the Italian data protection authority, the Garante, in its order against Open AI on ChatGPT. Although Open AI has complied with the requests from the Garante and ChatGPT is back live in Italy, companies developing and exploiting artificial intelligence risk facing strong headwinds with regulators taking inconsistent views on compliance requirements. In this sense, the EDPB’s task force on ChatGPT will hopefully ensure some level of consistency.
Overall, the provisional agreement of the EU Parliament on the AI Act is good news as it will create greater legal certainty in a sector that has often been subject to contradictory and overly protectionist positions. Businesses planning to incorporate AI into their operations and customer services must already take it into account to ensure their artificial intelligence strategy is future-proof. This assessment should be done in conjunction with the analysis of the impact of IP, privacy, and cybersecurity rules, as well as technical standards like ISOs, which regulators often use to spell out legal requirements.
Businesses cannot afford to miss the AI train, but they need to get prepared for the regulations and potential challenges ahead.
On a similar topic, you can find the following article interesting “ChatGPT will have stricter rules under the EU AI Act as foundation models“.
Image from the book The AI Revolution by William Harris