Have you already used the famous ChatGPT for work or to look up information for personal reasons? If you have, you’ll have noticed that its speed and efficiency are as fascinating as they are unsettling. If, whilst still in its ‘infancy’, this technology is capable of producing such results, just imagine what it will be like in a few years’ time, when a single tool will be able to do all the work and the market will become a ‘red ocean’ full of sharks competing for the AI oligopoly.
To curb this almost apocalyptic situation that is looming, and bearing in mind the associated problems that are arising in relation to privacy, data and the loss of many existing jobs, the EU has drawn up a draft law to regulate the development of AI.
What is generative AI, and how does it differ from artificial intelligence?
Before delving into the underlying principles and reasons behind this proposed legislation, it is important that we first clarify the difference between conventional artificial intelligence and generative artificial intelligence, as it is the latter that has sparked the controversy.
The artificial intelligence focuses on the analysis and interpretation of data, whilst generative artificial intelligence can create original and unique content from scratch. The fact that this technology has burst onto the scene with its creative dimension represents a truly unexpected breakthrough, but also a threat, since, until now, we all felt safer behind the barrier that human creativity placed between technology and people. Well, that barrier no longer exists.
Artificial Intelligence Act: different rules for each risk level
To shed light on this uncertainty that has spread amongst businesses and users worldwide, the European Union aims to protect its members from data theft and online privacy breaches through the new Artificial Intelligence Act. Would you like to know what it involves? We’ll explain it to you step by step below!
The priority of this law is to ensure that AI systems used within the EU are safe, traceable, transparent and inclusive. To this end, AI systems are analysed and classified according to the risk they pose to users, and for the different levels of risk, regulations of varying degrees of flexibility will be developed, categorising them as: unacceptable risk, high risk, generative AI and limited risk. Once approved, these will be the world’s first regulations on AI, so let’s take a look at what each category entails.
What are the different types of AI risks covered by the new Artificial Intelligence Act?
Unacceptable risk: AI systems posing an unacceptable risk are considered a direct threat to individuals and will be banned if they involve: cognitive manipulation of the behaviour of vulnerable individuals or groups; and the classification of individuals based on their behaviour, status or personal characteristics. This also includes real-time biometric identification systems.
High risk: these are all AI systems that infringe fundamental rights, and they are divided into two groups: those AI systems used in products subject to EU product safety legislation and the AI systems in the following eight areas, which must be recorded in a database:
- Biometric identification and categorisation of individuals.
- Management and operation of critical infrastructure.
- Education and vocational training.
- Employment, self-employment and workforce management.
- Access to and enjoyment of essential private services and public benefits.
- Enforcement of the law.
- Immigration management, border control and refugee asylum.
- Assistance with legal interpretation.
Generative AI: Generative AI systems will be permitted to continue operating provided they meet the following requirements:
- To ensure that the content has been created by AI.
- You design a model that prevents illegal content.
- Publish summaries of data protected by copyright.
Limited risk: these are the ones that meet certain minimum requirements, and users will always be fully informed. For example, they must be informed if the content they are viewing is a deepfake or has been manipulated.
Certain practices prohibited under the Artificial Intelligence Act
Case law on artificial intelligence may vary from country to country, but there are nevertheless certain prohibited practices that are common to all legal systems, such as the following:
Unfair discrimination: The law may prohibit the use of artificial intelligence algorithms that perpetuate discrimination based on characteristics such as race, gender, religion or sexual orientation. This includes automated decisions in areas such as recruitment, credit and the criminal justice system.
Mass surveillance without consent: the law may impose restrictions on the use of technologies the use of artificial intelligence for the mass surveillance of individuals without their consent. This includes the recording or tracking of individuals without a sound legal basis or without adequate safeguards to protect their privacy.
Manipulation of information: The use of artificial intelligence to deliberately spread false or misleading information with the aim of influencing people’s opinions or actions may be prohibited. This includes the manipulation of recommendation algorithms or the creation of profiles that generate filter bubbles or information biases.
Risks to safety and human life: the Act will lay down specific regulations to ensure the safety and minimise the risks associated with the use of artificial intelligence systems. This may include banning autonomous systems that pose an unacceptable risk to human life or requiring appropriate safeguards in sectors such as autonomous transport or AI-assisted healthcare.
If your company is already one of the pioneers in using AI for its operations, or if you simply believe it could represent a major step forward for your workflow or that of your team, don’t hesitate to subscribe to Educa.Pro, where you’ll find all sorts of related training and information. We look forward to seeing you there!
