The European Parliament is about to cement its stance on the world’s first AI Act, with lawmakers reaching a provisional political agreement after weeks of debate. After discussions between the rapporteurs and shadow rapporteurs of the law, a vote will take place at the committee on May 11, 2023, followed by a plenary vote in June. If successful, it will see talks with the Council, the member states of the EU, to bring the law into effect before the end of the year.
The AI Act, as proposed by the European Commission, is comprehensive and regulates a wide range of AI applications, with the goal of promoting public trust through transparency and protecting rights and freedoms. It has also been designed to ensure safety and enforce ethical principles in the development and deployment of these AI systems.
However, the most discussed parts of the proposed regulations were those concerning generative AI, such as ChatGPT, DALL-E, and Midjourney. These systems are general-purpose and can handle a range of tasks, which were not allowed in the original draft. After the disruption caused by ChatGPT and other AI models, EU lawmakers agreed to regulate them with the goal of upholding fundamental rights and freedoms. Among other changes, it has been mandated that companies must disclose any copyrighted material used in the development of these systems.
Other than generative AI, the AI Act follows a classification system, grouping AI systems into four risk levels beginning with unacceptable and ending at minimal risk. Low and minimal risk systems, like spam filters and video games, may be used but with some transparency, while systems posing a serious risk must comply with regulations that involve testing and human oversight.
The final text of the agreement will be the result of minor adjustments and the approval of all members of the EU Parliament. It is of paramount importance to ensure this law is effective and in place to protect the public from the misuse of AI technologies.
Chris Steffen is an analyst at Enterprise Management. He remarks on the difficulty of regulating generative AI such as ChatGPT, which leaves little room for regulation without affecting their functionality. He mentions the potential for misuse of the technology and the challenges posed by the large language models (LLMs).
Overall, it is clear that the European Union is taking a decisive step to ensure that AI technologies are governed responsibly. The changes proposed by the AI Act demonstrate a caring attitude towards making sure humans are not taken advantage of while also taking into account any risks posed to rights and freedoms. For now, the EU awaits the results of the committee vote on May 11, 2023, and hopes that talks with the Council will be successful in ensuring the enactment of the AI Act.
