Does the EU AI Act Apply to Your Business?

Does your startup rely on LLMs or NLP? The European Parliament has officially approved the European AI Act, and the Law came into force in 2024.

This article has been updated and written based on our Founder's article with Sparring.

In a landmark decision in 2023, the European Parliament established its stance on AI regulation. The Act itself was designed designed to mitigate risks associated with AI technologies while ensuring their ethical and responsible use for European consumers.

Looking to find out if the EU AI Act applies to your company? Get your free assessment here.

Assenteo - Agreeable Data Compliance for Fast Growth Businesseswww.assenteo.com

If your company integrates AI into its tech stack or builds revenue models around AI, here are the key takeaways you need to know.

1. Is the technology you are creating a prohibited or high risk AI system?

The AI Act follows a risk-based approach, banning AI applications that pose harm to people’s safety or are discriminatory. Certain systems that are strictly prohibited include those that ‘deploy subliminal or purposefully manipulative techniques, exploit people’s vulnerabilities or are used for social scoring (classifying people based on their social behaviour, socio-economic status, personal characteristics)’. Furthermore, individuals will have more rights to be protected from the negative risks of using ‘high-risk’ systems, including AI that handles people’s health, safety, fundamental rights, the environment, influencing voters in political campaigns and in recommendation systems used by social media platforms.

By taking this approach, the European Parliament aims to safeguard individuals and organizations from the adverse effects of AI systems. Furthermore, it should provide certainty for businesses. It will however, result in certain additional disclosures and documentation for startups who are creating technology in specific verticals such as healthcare.

2. Does your technology use Remote Biometric Identification?

During the parliamentary committee-level discussions, one of the main points of contention was the real-time use of Remote Biometric Identification. While some wanted exceptions for specific circumstances like terrorist attacks or locating missing individuals, the ban on real-time use mostly prevailed.

3. Do you utilise generative AI?

Perhaps the question most influential to societal AI use: how does the AI Act impact generative AI? The European Parliament introduced a tiered approach for AI models, with a focus on foundation models and generative AI systems.

To address concerns regarding transparency and accountability, the European Parliament has proposed mandatory labeling for AI-generated content and the disclosure of generative AI models containing training data protected by copyright. This requirement will apply to models such as ChatGPT and Bard for example.

4. Are you developing models?

Providers and developers of foundation models are to be held to a higher standard and would need to assess if their model falls within a high-risk category. Builders of models should also be aware they will need to register their models in the EU database prior to release on the EU market.

Unsure if the EU AI Act applies to you? Get your free assessment here.

Assenteo - Agreeable Data Compliance for Fast Growth Businesseswww.assenteo.com

---