ChatGPT, Gemini, Grok… Europe’s Tough New AI Rules Are Now in Force

On August 2, 2024, the European Union entered a new era of tech regulation with the enforcement of the AI Act—a landmark law hailed as the most comprehensive AI framework in the world. Its mission: to bring transparency, safety, and fundamental rights protection to some of the most powerful AI models, including GPT-4, Gemini, and Grok.

While Brussels sees the legislation as a tool to strengthen digital sovereignty and safeguard democratic values, not everyone is happy. The United States and several tech giants warn that these rules could slow innovation and undermine global competitiveness.

Immediate Obligations for Generative AI Models

The AI Act imposes strict requirements on providers of General Purpose AI (GPAI) models capable of generating text, images, or videos. These companies must now:

• Provide a full technical documentation explaining model design and functionality.
• Publish a public summary of training data sources.
• Implement a clear copyright compliance policy.
• For models classified as systemic risk, notify the European Commission, conduct risk assessments, and adopt enhanced security measures.

The strictest rules apply immediately to the most advanced AI models, while others have until August 2, 2027 to comply.

A Transatlantic Tension Point

Not surprisingly, the AI Act has sparked political friction:

• Washington has criticized the approach as “overly restrictive,” launching a National AI Acceleration Plan to cut red tape and maintain U.S. leadership.
• Google, while signing the EU’s AI Code of Practice, has expressed concern the rules could slow down AI deployment in Europe.
• Meta refused to sign, citing “too many uncertainties” in the regulation.

Despite the pushback, the law is legally binding, with fines of up to 7% of global turnover for violations.

The AI Act: A World-First Legal Framework

Adopted in March 2024, the regulation (EU) 2024/1689 establishes the world’s first harmonized framework for AI development, commercialization, and use.

It follows a risk-based approach, classifying AI systems into four categories:

1. Unacceptable Risk – outright banned (e.g., social scoring, real-time remote biometric identification in public spaces, exploiting vulnerabilities).
2. High Risk – strict oversight for AI in biometric ID, education, employment, infrastructure, and justice.
3. Limited Risk – must inform users they are interacting with AI.
4. Minimal or No Risk – no specific regulatory obligations.

Key Implementation Timeline

• August 1, 2024 – AI Act enters into force.
• February 2, 2025 – Ban on unacceptable-risk AI systems.
• August 2, 2025 – Rules for general-purpose AI models and designation of national authorities.
• August 2, 2026 – Enforcement for existing high-risk AI systems and launch of regulatory sandboxes.
• August 2, 2027 – Full application to regulated products embedding high-risk AI.

Regulatory Sandboxes: Encouraging Innovation

Regulatory sandboxes are supervised environments allowing companies—especially SMEs—to test AI systems in real-world conditions while benefiting from temporary regulatory flexibility.
Advantages include:

• Controlled risk evaluation.
• Faster product adaptation to compliance requirements.
• Reduced market-entry obstacles.

Impact on Businesses

Companies developing or deploying high-risk AI systems will have to:

• Obtain CE marking.
• Register in the EU AI database.
• Implement risk management and data governance frameworks.
• Guarantee traceability, transparency, and human oversight.
• Ensure robustness, accuracy, and cybersecurity.

From August 2025, general-purpose AI model providers must:

• Increase transparency.
• Comply with strict copyright protection.
• Meet reinforced obligations for systemic-risk models.

A Strategic Vision for Europe

In France, the AI Act is seen as a chance to strengthen innovation and develop homegrown expertise. The government has committed €400 million to AI hubs, aiming to train 100,000 specialists annually.

With this regulation, the EU seeks to balance innovation with the protection of citizens—an approach that could either secure Europe’s place in the global AI race or risk leaving it behind faster-moving competitors.

Conclusion

The AI Act positions Europe as a global leader in AI regulation, setting a precedent for how technology and ethics can coexist. But with mounting pressure from industry and global rivals, the coming years will reveal whether Europe’s balance between control and innovation becomes its competitive advantage—or its Achilles’ heel.