EU expands on 'AI' definition under act

The European Commission has published guidelines aimed at helping providers and others to determine whether a software system constitutes an artificial-intelligence (AI) system under the EU’s AI Act.

The guidelines, published last week, are not binding, however, and the commission says that they are designed to “evolve over time”.

The commission adds that any binding interpretation of the EU’s AI legislation can be given only by the Court of Justice of the European Union.

Under the AI Act, AI systems are classified into different risk categories, with those at the highest level of risk prohibited, and higher-risk system subject to stricter regulation.

The first rules under the act started to apply on 2 February.

Seven elements

Under the EU legislation an ‘AI system’ is defined as:

“a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments.”

The commission guidelines break down the seven main elements of the definition:

1)    A machine-based system,
2)    That is designed to operate with varying levels of autonomy,
3)    That may exhibit adaptiveness after deployment,
4)    And that, for explicit or implicit objectives,
5)    Infers, from the input it receives, how to generate outputs,
6)    Such as predictions, content, recommendations, or decisions,
7)    That can influence physical or virtual environments.

The commission stresses that the definition of an AI system adopts a “lifecycle-based perspective” encompassing two main phases:

• The pre-deployment or ‘building’ phase of the system, and
• The post-deployment or ‘use’ phase of the system.

It adds that the seven elements set out in the definition are not required to be present continuously throughout both phases of that lifecycle – specific elements may appear at one phase, but may not persist across both phases.

The commission says that this approach reflects “the complexity and diversity of AI systems”, and ensures that the definition aligns with the act's objectives by accommodating a wide range of AI systems.

‘Gigafactories’ plan

The commission guidelines were published ahead of a summit on AI being held in Paris, at which commission president Ursula von der Leyen today (11 February) launched a €200 billion public-private investment programme called InvestAI.

It includes a fund of €20 billion to finance four AI ‘gigafactories’ across the EU. These will specialise in training the most complex, very large, AI models.

“This unique public-private partnership, akin to a CERN for AI, will enable all our scientists and companies – not just the biggest - to develop the most advanced very large models needed to make Europe an AI continent,” said von der Leyen.