GPT: The European Parliament and the Council have recently reached a provisional agreement on the Artificial Intelligence Act, a significant regulatory framework for the use of artificial intelligence (AI) in the European Union. The agreement aims to ensure that AI systems in the EU are safe and respect fundamental rights, while also fostering innovation and positioning Europe as a leader in the field. Key aspects of the Act include safeguards on general purpose AI, limitations on the use of biometric identification systems by law enforcement, bans on social scoring and AI used to manipulate or exploit vulnerabilities, and rights for consumers to lodge complaints and receive explanations. Specifically, the Act prohibits certain applications of AI that pose a threat to citizens' rights and democracy. These include biometric categorisation systems using sensitive characteristics, untargeted scraping of facial images, emotion recognition in workplaces and educational institutions, social scoring based on social behaviour, and AI systems that manipulate human behaviour or exploit vulnerabilities. For high-risk AI systems, the Act establishes clear obligations, including a mandatory fundamental rights impact assessment. This applies also to the insurance and banking sectors, and AI systems used in influencing elections and voter behaviour are classified as high-risk. The Act also sets out obligations for high-impact general-purpose AI systems, which include risk assessments, adversarial testing, incident reports, and transparency requirements. In terms of enforcement, non-compliance with the rules can result in fines ranging from 35 million euros or 7% of global turnover to 7.5 million euros or 1.5% of turnover, depending on the infringement and the size of the company. The agreed text will now need formal adoption by both the Parliament and Council to become EU law, with the Parliament’s Internal Market and Civil Liberties committees set to vote on the agreement in an upcoming meeting. The legislation is not expected to take effect until at least 2025. For more detailed information, you can refer to the articles from the European Parliament's news page [here](6) and from BNN [here](7). Me: yes, but consider the following passage: For high-impact GPAI models with systemic risk, Parliament negotiators managed to secure more stringent obligations. If these models meet certain criteria they will have to conduct model evaluations, assess and mitigate systemic risks, conduct adversarial testing, report to the Commission on serious incidents, ensure cybersecurity and report on their energy efficiency. MEPs also insisted that, until harmonised EU standards are published, GPAIs with systemic risk may rely on codes of practice to comply with the regulation. What 'codes of practice' means, in practice? Him: In the context of the European AI Act, "codes of practice" refer to a set of guidelines or standards that organizations can follow to ensure compliance with the regulation, particularly for general-purpose AI (GPAI) systems that pose systemic risks. These codes are intended to provide practical guidance on how to meet the Act's requirements, such as conducting model evaluations, assessing and mitigating systemic risks, performing adversarial testing, ensuring cybersecurity, and reporting on energy efficiency. Until harmonized EU standards are established, these codes of practice offer a framework for organizations to align their AI systems with the Act's objectives and principles. For more detailed information, you can refer to the European Parliament's website: [Artificial Intelligence Act: deal on comprehensive rules for trustworthy AI]( https://www.europarl.europa.eu/news/en/press-room/20231206IPR15699/artificia... ). Me: Please help me understand. What 'regulatory sandbox' could mean in the following passage? MEPs wanted to ensure that businesses, especially SMEs, can develop AI solutions without undue pressure from industry giants controlling the value chain. To this end, the agreement promotes so-called regulatory sandboxes and real-world-testing, established by national authorities to develop and train innovative AI before placement on the market Him: A "regulatory sandbox" in this context refers to a controlled environment where businesses, particularly small and medium-sized enterprises (SMEs), can develop and test innovative AI solutions under relaxed regulatory conditions. This setup allows for real-world testing and development without the full burden of compliance with extensive regulations that might normally apply. These sandboxes are established by national authorities to encourage innovation in AI while ensuring safety and compliance standards are met before these AI solutions are introduced to the broader market. For more detailed information, you can refer to the European Parliament's website: [Artificial Intelligence Act: deal on comprehensive rules for trustworthy AI]( https://www.europarl.europa.eu/news/en/press-room/20231206IPR15699/artificia... ) Me: Suppose I'm a SME and I'm in the process of developing an AI solution based on LLM. Why should I care about a “regulatory sandbox”? Him: As an SME developing an AI solution based on Large Language Models (LLM), a regulatory sandbox offers significant advantages. It allows you to test and refine your AI product in a real-world environment with reduced regulatory constraints. This means you can innovate and experiment more freely, accelerating development while ensuring compliance with essential safety and ethical standards. Essentially, a regulatory sandbox provides a safer, more flexible space for you to bring your AI innovations to market more efficiently and effectively. For more details, visit the European Parliament's website: [Artificial Intelligence Act: deal on comprehensive rules for trustworthy AI]( https://www.europarl.europa.eu/news/en/press-room/20231206IPR15699/artificia... ). Tutto a posto, buona domenica! Guido