The EU AI Act: How Other Countries Are Regulating AI

The European Union passed comprehensive artificial intelligence legislation, establishing the regulatory framework for AI use within the EU. The laws are intended to protect individual fundamental rights, support AI innovation by EU start-ups, and improve state-led conditions for AI development.

The EU’s AI Act uses a risk-based approach to apply different standards to AI based on the potential risk factors. The EU law’s broad reach means the Brussels Effect will extend to other AI providers around the world. To stay up-to-date with AI regulations in the EU and what they mean for U.S. companies, talk to an experienced science and technology lawyer.

What Is the EU AI Act?

The Artificial Intelligence Act is a European Union regulation establishing basic standards for AI. The EU’s AI Act entered into force on August 1, 2024. 

The legal framework applies to most AI governance among stakeholders, providers, and systems using AI technologies. However, there are different legal regulations and codes of practice based on the level of risk associated with the use of AI applications.

Find top Science and Technology Law lawyers easily

Connect with a qualified attorney today.

Find a lawyer today

Understanding the Risk-Based Approach to Regulation

The European AI Act uses a risk-based approach to regulate AI applications, based on the perceived risks to fundamental rights. The AI risk management categories include:

• Unacceptable risk AI
• High-risk AI systems
• General-purpose AI models
• Limited risk AI practices
• Minimal risk

Unacceptable risks are banned in the EU. Prohibited AI practices include social scoring, emotion recognition in workplaces or schools, the untargeted scraping of facial images to build databases, and the use of “real-time” remote biometric identification systems in publicly accessible spaces for law enforcement. However, strictly defined exceptions exist for real-time identification in cases involving terrorism or searching for missing persons. 

High-risk AI applications pose serious threats to a person’s fundamental rights, health, or safety. This can require additional transparency obligations and human oversight. High-risk AI systems include those that fall under EU product safety legislation and EU database registration systems (including health, education, and critical infrastructure management).

Generative AI models (like ChatGPT) are regulated under a specific framework that imposes stricter obligations on models classified as having systemic risks. This includes disclosure that output is AI generated. General purpose AI with the potential for systemic risk is still subject to review, technical documentation, and evaluation.

Even limited risk AI systems have transparency obligations, including disclosure about AI use. Minimal risk categories are not regulated. However, other EU and member state laws may apply to companies providing minimal risk AI services.

Responsible AI Development

Another focus of the EU AI Act is to help small and medium-sized enterprises (SMEs) develop and compete in the AI marketplace. Large global corporations have access to money and large data sets to train their AI foundation models. However, large tech companies can pursue development at the cost of fair competition and the protection of fundamental rights.

As an alternative, the EU regulations encourage trustworthy AI based on quality data, real-world testing, and proper documentation. Member states work with start-ups and SMEs to create and train new algorithmic models for developing AI. For example, the EU’s AI Innovation Package provides support to smaller AI start-ups, including use of EU supercomputers.

How EU Law Shapes Global Standards

EU regulations apply to multiple EU member states, millions of people, and major corporations. AI deployers follow EU standards to avoid noncompliance with a large share of the market. AI providers following EU regulations often adapt those same standards to other jurisdictions (like the EU’s General Data Protection Regulation).

Response to the EU AI Act

After the EU AI Act, the Biden Administration developed a blueprint for an AI Bill of Rights to create a framework for general AI regulations and accountability. The Executive Order established the following five guiding principles:

• Safe and effective systems
• Algorithmic discrimination protections
• Data privacy
• Notice and explanation
• Human alternatives, consideration, and fallback

No comprehensive legislation has been passed by Congress as of this time. While there is currently no single comprehensive federal AI law, agencies like the FTC and DOJ actively regulate AI development using existing consumer protection and civil rights authorities.

EU AI Regulations and Legal Advice for U.S. Companies

The broad reach of the EU’s AI law can affect the start-ups and AI providers in the U.S. Companies offering AI services to users in the EU must comply with the EU AI laws, or face noncompliance fines and penalties. The EU AI Act establishes administrative fines of up to 35 million euros or 7% of global annual turnover, whichever is higher, for noncompliance with prohibited practices.

Following EU AI legislation also provides a broader market for companies innovating AI services. For legal advice about AI compliance in the EU, U.S., and other jurisdictions, contact an experienced science and technology attorney.