Umanis Supports Its Clients in Promulgating the “Artificial Intelligence Law”

Umanis Supports Its Clients in Promulgating the "Artificial Intelligence Law"

[ad_1]

Umanis (www.umanis.com), the French leader in data, digital transformation, infrastructure and cloud, is already working on new European regulations related to artificial intelligence. This should see the light of day in 2022.

A group of experts commissioned by the European Commission has assessed that even the most innovative and replicable AI techniques cannot guarantee that there are no flaws or weaknesses. The use of artificial intelligence is constantly evolving in the daily life of the company and the user. Experts have noted that systems are already abundant and difficult to implement and thus difficult to control. AI often accompanies large amounts of personal data that can harm individuals. The use of these tools can lead to discrimination that, depending on their use, can harm different groups of people: prejudices of a sexual and ethnic nature, predictions of the frequency of criminal acts, etc.

The European Commission, based on the recommendations of this group of experts, has decided that it is necessary to create or rebuild trust in AI without impeding its innovations. The European Parliament decided to regulate this technology by implementing the AIA (Artificial Intelligence Act) which aims to harmonize the rules related to artificial intelligence systems.

Actors involved by the IAA

This regulation defines three main categories of actors: suppliers, distributors and users established in the European Union or who have established an AI system on the territory of the European Union.

The supplier is the main actor. You must propose and maintain the operation of an AI system that complies with the regulations. He undertakes to do so by issuing a declaration in a dedicated European registry.

The distributor and the user are each responsible when distributing or using an incompatible, unregistered, and undocumented artificial intelligence system…

Artificial intelligence systems concerned with this regulation

Not all AI systems are affected. The regulation defines 4 categories of artificial intelligence systems:

Forbidden AI systems are high risk, AI systems with acceptable risks, and AI systems that are less dangerous. It is up to the actors to prove and document that their AI systems belong to one category or another. The first urgency will be to prove that they do not belong to the first category: banned AI systems. The regulations apply to Class II: High-Risk AI systems.

Its implementation will lead to many commitments such as documenting uses, integrating cybersecurity devices, certification of strength, transparency about the AI ​​systems used, and establishing human oversight. This regulation is directly applicable in the legal system of member states. It is binding on all subjects of law: individuals, legal persons, states, institutions.

Penalties incurred

Failure to comply with this regulation will result in penalties 24 months after the date of publication. Violators are fined up to 6% of their annual global turnover or 30 million (whichever is greater).

Regulations Basics

This regulation is based on a risk-based approach. Compromise with GDPR can be done based on the same approach.

A few keys to compliance

The AIA sets out 7 core principles (compliance checkpoint subject to sanction):

1) human factor and human control (tracking),

2) Technical durability and safety,

3) privacy and data management,

4) Transparency,

5) Diversity, non-discrimination and fairness,

6) Societal and environmental well-being, and

7) accountability.

The European legislator estimates the time required for compliance to be 24 months. Feedback from compliance projects with other binding regulations encourages us to encourage anticipation. Umanis wants to provide its customers with high-level support to ensure compliance with European AI requirements,” says Patricia Chemali-Noel, Director of GDPR Practices “Umanis has developed a structured and comprehensive approach to bring simplicity and smoothness in the business operations of its customers.”

To respond to this regulation, which can sometimes seem complicated, Umanis has developed an approach based on several axes:

Risk management and reduction :
The company must identify, plan and qualify its AI systems and manage, allocate and reduce known, expected or potential risks.

Announcing the artificial intelligence system(s) :
The record must indicate all high-risk AI applications including autonomous AI systems. Anticipation will make it possible to collect the required documents from other actors identified in the previous stage.

keep a record :
Compliance and risk management documents must be maintained. These compliance documents are based on a commitment to transparency, providing information to users, and establishing human oversight over artificial intelligence. This document must also provide the quality, robustness, accuracy, and security of its AI system.

Data Governance Department :
The company must demonstrate that its data governance is in line with AIA requirements. In order to ensure control over data accuracy, system robustness as well as protection from cyber attacks, the company must justify the implementation of effective operations. Finally, reliability tests must be relevant, representative, error-free and complete.

Create an accident reporting process :
The Provider and the User are obligated to inform the relevant national authorities of any serious accident or malfunction that constitutes a violation of obligations in terms of the fundamental rights of citizens.

[ad_2]

Source link