On April 21, 2021, the European Commission unveiled the Artificial Intelligence Act, a regulatory project aimed at establishing a common regulatory and legal framework for artificial intelligence (AI) and its uses within the European Union (EU).
The objectives of this initiative are manifold: to impose a regulatory framework to oversee the use and development of AI, strengthen governance in this field, and attract investments through a robust legal framework aligned with global developments.
This regulatory project applies to AI providers and distributors operating in the EU, AI users within the EU, and providers and users of AI systems located in a third country where the solution produced by these systems is used in the European Union.
In this article, we will address the various rules defined for the development of artificial intelligence systems, Europe’s positioning in this dynamic, and how SaaS solutions like T.O.P. align with the ethical approach in compliance with the AI Act.
Rules Defined from the Inception of AI Systems:
The AI Act obliges companies to meet certain requirements in the development of their AI systems:
- Human Oversight: Companies must establish human supervision of AI to control the actions of AI systems.
- Risk Management: Anticipating, identifying, and eliminating risks is crucial to reducing risks associated with AI use in their systems.
- Data and Data Governance: Rigorously test and validate all data used for AI to ensure optimal governance.
- Transparency and User Information: Companies must allow an interpretation of results understandable to all and inform users transparently about the use of their data.
- Traceability: Events related to AI must be automatically recorded to ensure adequate traceability.
- Documentation: Provide documentation established before market release to assess AI compliance.
- Accuracy, Robustness, and Cybersecurity: Companies must meet and achieve the goals of these aspects throughout their life cycle.
Classification Matrix for AI Solutions:
The AI Act will define a risk classification matrix for these solutions:
- Systems Considered Unacceptable Risk:
- Social Scoring: AI cannot socially score a person due to risks of biases, discrimination, and privacy violation.
- Biometric Identification: AI cannot instantly identify a person using biometric data.
- Subliminal Techniques: AI cannot use harmful subliminal techniques to influence thoughts, emotions, and behaviors subconsciously without individuals’ consent.
- Exploitation of Vulnerable Individuals: Exploiting vulnerable individuals, such as the elderly, disabled, or children, using AI is morally reprehensible.
- Systems Considered High Risk: These include solutions representing predictable dangers to health, safety, fundamental rights, the environment, democracy, or the rule of law.
- Systems Considered Medium or Limited Risk: These are systems that interact with humans, detect emotions, determine association with social categories based on biometric data, or generate and manipulate “deep fakes.”
- Systems Considered Minimal Risk: Examples include anti-spam filters or video games that do not require specific obligations and pose no real dangers.
How Europe Fits into the AI Development Dynamic?
The AI Act responds to a global dynamic on Artificial Intelligence: market growth and numerous challenges to exploit its potential ethically and beneficially for humanity.
In a broader sense, this regulation and support for AI growth are still underrepresented in the United States, as they lack a federal framework to regulate it. The Algorithmic Accountability Act has been introduced in the U.S. Senate and the House of Representatives, but even if adopted, it will not take effect for several years.
However, some states like Alabama, Colorado, Illinois, or Mississippi have already enacted laws in favor of AI regulation.
In this sector, Europe positions itself as a good example of support and structuring on the subject of regulation through the establishment of several institutions and regulations:
- ECAT (European Centre for Algorithmic Transparency): Provides technical expertise to ensure that large online platforms and search engines comply with risk management, reduction, and transparency rules related to digital services legislation.
- GDPR (General Data Protection Regulation): A European regulatory text that protects and regulates data processing equally in the European Union.
- CNIL (Commission nationale de l’informatique et des libertés): An independent authority that monitors and controls the use of information technology to ensure compliance with French law, including the protection of personal data and human rights.
- LaborIA: Created by the Ministry of Labor and INRIA in November 2021, aims to understand artificial intelligence and its impact on work, employment, and skills.
- OECD (Organisation for Economic Co-operation and Development): Studies the economy and aims to encourage public policies for prosperity, equal opportunities, and the well-being of all. This organization is also interested in various AI developments.
How TOP Supports Ethical Artificial Intelligence Aligned with the AI Act?
From its inception, TOP has embraced ethical and responsible data usage. Since the beginning of its development, TOP experts have ensured alignment with the AI Act’s criteria.
As a decision support SaaS solution, TOP values human resources management by advising managers or HR on how to anticipate and avoid employee resignations. The solution user, the manager, retains control over the advice provided by the solution while remaining the sole decision-maker.
To prevent influencing results through stigmatizing or unfair social biases, TOP has implemented debiasing mechanisms in its algorithmic results. The goal is to ensure fairness and impartiality in predictions provided by TOP.
To ensure transparency and compliance with its AI system, TOP is monitored by Hub France AI and Lab RH, organizations specialized in AI ethics and regulation.
Compliance with CNIL standards is also a priority for TOP, aligning with regulations defined by GDPR, including hosting its databases and backups in Europe.
TOP addresses data processing challenges from the users’ perspective. When the data registry requires it, their teams provide users with clear information about the use of their data and have them sign terms of use.
To guarantee security, the solution encrypts all data, preserving the confidentiality and integrity of information. TOP also records all events to have complete traceability, maintaining a history of all actions taken and results obtained.
All actions and results generated by TOP are intended to serve the employee in their relationship with the company, aiming to provide added value to each individual by offering reliable information and relevant recommendations to facilitate decision-making.
The AI Act, the European regulation on artificial intelligence, aligns with the global trend to place the European Union on par with the rest of the world. Europe appears advanced on many points, particularly on the security and legal framework for the use of these technologies. TOP, since its inception, has followed all these dynamics to propose an ethical model in line with the upcoming rules.
