AI should be human centric, the ultimate goal for its development shall be human centric and the thought that AI is pecuniary. With this hopeful idea (and assumption), the governments of the world are even allowing the AI markets in their country for their citizen. The ultimate reason for any country to allow AI market is for its own economic and societal benefits. Without any debate, almost everyone agrees that legislature should step into it and make laws to regulate the AI. Because we generally think that if anything is done under the roof of law is beneficial (until you know a lawyer), and ultimately we feel safe with that.

So, for that, purposes European Union (EU) had enforced the EU AI act on August 1, 2024; before that European Parliament (EP) had also taken considerable amount of action. For instance, in October 2020 EP adopted numbers of resolution related to AI which covered areas like ethics, liability and copyright. In 2021 resolutions were also adopted regarding criminal matters, education, culture and the audio-visual sector. These resolutions governed development, deployment and use of the AI.

The law is tailored with the risk-based approach and built on 4 specific objectives:

• AI systems that are placed in markets are safe and respectful toward the law and fundamental rights and values of the EU.
• To foster investment and innovation in AI with clearly laid down legal principles.
• Effectively implementing the existing laws on fundamental rights and enhance governance on AI safety systems.
• Prevent market fragmentation of AI and facilitate trustworthy, lawful and safe AI applications.

According to the explanatory memorandum of the AI act, legal interventions are made where there is a justified cause for concern or where such concern can reasonably be anticipate in near future. There is a well-defined risk based regulatory approach that doesn't make a trade corrosive by putting unnecessary restrictions. For an aeon the EU AI act accentuates on problem faced while developing of an AI system and use of the same system.

The proposal defines common mandatory requirements applicable to design and development of AI. The primary objective of the act is to ensure the stability of internal market of the (European) Union.  Not just the union who is concerned about the risk of AI or in future but the Nations in the union is also developing their domestic laws and regulation, for safety of its own citizen, which is a good thing. But, there is other side to that also; with that many laws a legal uncertainty will arise which will cumbersome the pace of AI uptake in the market. That too will create a fragmentation of the internal market in EU regarding essential elements of a particular AI system. This legal bewilderment will certainly arise for both users and providers of AI systems.

When talking about act itself, it laid down the definition of AI and makes sure it also adaptable for the near future possible developments as AI is fast developing condominium. The participants in the market chain of AI are also well defined, at both public and private level. Following that the act established a list of prohibited AI(s), which is follows a risk based approach. Than this risk based approach divides the AI systems from an unacceptable risk to minimal risk. The law is well established on 'unacceptable risk AI models', but it more concerned about 'high-risk AI models', so that TITLE III is epically dedicated to it , it's more like Union want to make this topic an ostentatious.

I think (European) Union assumed that the 'unacceptable risk AI' is way far in the future and they need to focus on which is in near future or is currently present (and there is nothing wrong in it.). For that to happen the providers of the AI and legislature have to develop a significant amount of trust on each other. But at the top of it the Act would majorly be dependent on the implementation of the same, for the title(s) VI,VII and VIII, will look upon the governance and implementation part.

Title VI will establish governance system at Union and National level, at union level the proposal establishes a 'European Artificial Intelligence Board', for the national level country members of the Union have to designate 1 or more national authorities. Title VII & Title VIII will be subservient for the monitoring work and reporting obligations for the providers of AI respectively.

Title I: General Provision

Title I consists of 4 articles (Article 4 contains amendments to annexure).

• Article 1: Lays down the purpose of the law, focusing on AI. It broadly states the topics considered while forming this law.
• Article 2: Covers the applicability of the law. Regulations do not apply to AI systems developed or used for military purposes.
• Article 3: Defines key terms. Definitions are crucial for clarity in regulations.

What Is AI (according to the Act)

AI is defined as software developed using one or more of the following techniques and approaches:

• Machine learning approaches: Includes supervised, unsupervised, and reinforcement learning, with methods such as deep learning.
• Logic- and knowledge-based approaches: Includes knowledge representation, inductive (logic) programming, knowledge bases, inference and deductive engines, symbolic reasoning, and expert systems.
• Statistical approaches: Includes Bayesian estimation, search, and optimization methods.

AI operates based on human-defined objectives and generates output or decisions influencing its environment.

Key Entities

• Provider: A natural or legal person, public authority, agency, or body that develops AI systems or markets them under their name or trademark.
• Importer: Acts as a connector between the Union and external entities. An importer is a natural or legal person established in the Union who places AI systems in the market.
• Distributor: A person distinct from the provider and importer, who makes AI available in the market without altering its properties.

Title II: Prohibited Artificial Intelligence Practices

Title II contains only one article:

Article 5: Prohibited AI Activities

• AI systems that cause a person to harm another (physically or psychologically) and exploit vulnerable groups.
• AI deployed by public authorities to assess social behavior or predict personal characteristics, leading to unjustified unfavorable treatment.
• Real-time remote biometric systems in public spaces, except in the following cases:
  • (a) Identifying potential crime victims, including missing children.
  • (b) Preventing specific or imminent threats to life, physical safety, or terrorist attacks.
  • (c) Detecting perpetrators or suspects of criminal offenses punishable by at least three years of custodial sentence or detention.

The use of real –time identification system in publicly accessible spaces shall be granted by a judicial authority or by an independent administrative authority of Member state Title III
High - Risk Ai Systems

Classification Of Ai Systems As High-Risk
Chapter 1 contains total 2 articles , which focuses on defining the high risk AI systems or models. Total 8 AI systems have been listed as high risk . Covering the biometric identification , migration / asylum and border control , law enforcement , private and public service, employment and work management , education , management and operation of critical infra. AI systems that pose risk to health and safety, adverse impact on fundamental rights.
 
Requirements For High-Risk Ai Systems
This chapter incorporated for the risk management for high risk AI models. Containing total 7 articles. From risk management to accuracy, robustness and cybersecurity. Risk management includes identification, analysis, estimation and evaluation and reduction or elimination of High Risk AI, a process that will run through entire lifecycle of that AI.

Article 10 talks about the data and the governance, which further elaborates about testing , validation and testing data sets that only shall only take place for the intended purpose , relevant , representative , free of errors and complete. The technical documentation for these AI systems shall be drawn up before that system is placed in market and have to demonstrate complies. There shall be automatic recording of events with a level of traceability of AI.

Confidentiality And Penalties
National Authorities must respect the confidentially of information and data obtained. However , regardless this, information exchanged on a confidential basis between national authorities and commission shall not be disclosed without user and without prior consultation of the originating national authority. Meaning that these following instructions will be followed but it won't be affecting rights of the commission and member state of union and notified bodies with regard of information exchange.
 
Penalties
The member state shall lay down rules on penalties taking into an account the interest of small-scale providers and start-up and their economic viability.

Fines: infringing will be subjected to fines upto 30 000 000 euro or if the offender is a company the 6% of its total worldwide annual turnover ,whichever is higher.

Non-compliance of article 5(prohibited AI practices) and 10 ( data and governance) will lead to administrative fines upto 20 000 000 euro or, if offender is a company then 4% of its annual turnover for the preceding financial year, whichever is higher.

Providing incorrect , incomplete or misleading information to the notified bodies and national authorities ; the fine for that will be 10 000 000 euro , for companies , upto 2% of its total worldwide annual turnover , whichever is higher.

Conclusion
The EU AI Act represents a comprehensive and forward-looking legislative framework aimed at regulating the development, deployment, and use of artificial intelligence within the European Union. By adopting a risk-based approach, the Act seeks to balance the promotion of innovation and investment in AI with the need to ensure safety, respect for fundamental rights, and ethical considerations. The Act categorizes AI systems based on their potential risks, from unacceptable to minimal, with a particular focus on high-risk AI systems that could significantly impact health, safety, and fundamental rights.

Key provisions of the Act include the establishment of clear definitions for AI and its market participants, the prohibition of certain harmful AI practices, and the imposition of stringent requirements for high-risk AI systems. These requirements encompass risk management, data governance, technical documentation, and cybersecurity, ensuring that AI systems are developed and used responsibly. Additionally, the Act emphasizes the importance of transparency, accountability, and traceability in AI operations.

The governance structure outlined in the Act, including the creation of the European Artificial Intelligence Board and national authorities, aims to facilitate effective implementation and monitoring of AI systems. Penalties for non-compliance are substantial, reflecting the seriousness with which the EU regards the ethical and safe use of AI.

Overall, the EU AI Act is a significant step towards fostering a trustworthy AI ecosystem within the EU. By addressing potential risks and establishing clear legal principles, the Act aims to prevent market fragmentation, enhance consumer confidence, and promote the responsible use of AI technologies. The success of the Act will largely depend on its implementation and the ongoing collaboration between AI providers, regulators, and other stakeholders to ensure that AI development remains aligned with human-centric values and societal benefits.