In the new era of AI legislation, the European Union’s Artificial Intelligence Act (EU AI Act) is a foundational law regarding transparency rules for certain AI systems. It sets forth standards for AI utilisation across the member states, and amongst its intricate sections, Article 52 – focusing on transparency obligations for AI systems – holds particular significance. Let’s delve deeper.

AI system interaction: the emphasis on transparency (Art. 52.1)

An AI system under this Act should be designed such that individuals interacting with it can recognise its AI unless this is already apparent from the context. This stipulation delineates between AI and human interaction, helping build user trust in AI technologies.

However, specific AI systems employed for legal crime combat are exempt from this rule, barring those used by the public to report a crime.

Emotion recognition and biometric categorisation systems: ensuring accountability (Art. 52.2)

This Article accentuates the importance of transparency in AI systems that recognise emotions or categorise biometrics. As these systems permeate our lives, their potential to influence users – both positively and negatively – escalates. Individuals impacted by these systems must be aware of their operation, safeguarding their rights and privacy.

An example of this was the sentiment-analysis AI used by Budapest Bank without data subject consent, for which the relevant supervisory authority fined them under the GDPR.

However, systems legally utilised for crime investigations are an exception to this rule.

Deep fake AI systems: the pursuit of authenticity (Art. 52.3)

This Article focuses on deep fakes in a world grappling with misinformation. If AI is used to modify or create images, sounds, or videos to mirror reality, users must be informed of these alterations. Transparency is crucial in countering the potential harm deep fakes can cause.

There are exceptions in the context of legal action, freedom of expression, and the arts and sciences, providing sufficient protections for others’ rights and freedoms.

The harmony of new and existing regulatory requirements (Art. 52.4)

Article 52.4 underscores that the new obligations articulated in the rest of Article 52 do not supersede the requirements and obligations detailed elsewhere in the Act. This harmony ensures that advancements in AI don’t compromise the regulatory structures in place.

Grappling with the transparency rules for certain AI systems

The EU AI Act’s Article 52 paves the way towards a future of more transparent and accountable AI systems. It is a vital beacon in the landscape of AI law, with profound implications for companies, users, and society.

Actions you can take next