Europe’s World-First AI Rules Set for Final Approval
The AI Act represents a paradigm shift towards a human-centric approach to AI, where humans retain control over the technology.
European Executive Vice-President Margrethe Vestager (not seen) and European Commissioner in charge of internal market Thierry Breton hold a press conference on artificial intelligence (AI) following the weekly meeting of the EU Commission in Brussels on April 21, 2021. in Brussels on April 21, 2021. Photo by Anadolu Images.
E
uropean Union lawmakers are on the brink of giving their final approval to the much-anticipated Artificial Intelligence Act, marking a significant milestone in the regulation of AI technology globally. Scheduled for Wednesday, the impending vote in the European Parliament is poised to cement the EU’s position as a leader in AI regulation, with the rules expected to come into effect later this year.
The culmination of a five-year journey since its initial proposal, the AI Act is poised to shape the future landscape of AI governance. According to Dragos Tudorache, a Romanian lawmaker who played a pivotal role in negotiating the draft law, the AI Act represents a paradigm shift towards a human-centric approach to AI, where humans retain control over the technology to harness its potential for societal progress and economic growth.
Despite initial concerns and lobbying efforts from major tech companies, including OpenAI, the AI Act has garnered broad support. However, it’s worth noting that OpenAI CEO Sam Altman’s remarks last year about the possibility of withdrawing from Europe due to regulatory constraints sparked a brief controversy before being clarified.
Risks and usages
Risk-based approach: Similar to existing EU regulations, the AI Act adopts a risk-based approach to regulate AI applications. Systems deemed low-risk, such as content recommendation algorithms, will face lighter regulations, while high-risk applications, like those in medical devices or critical infrastructure, will be subject to stricter requirements.
Prohibited uses: Certain AI applications are outright banned due to their perceived risks, including social scoring systems, certain predictive policing techniques, and emotion recognition systems in educational and workplace settings. Additionally, the use of AI for remote biometric identification by law enforcement is restricted, except for specific cases like combating serious crimes.
Generative AI: The emergence of general-purpose AI models, such as OpenAI’s ChatGPT, prompted policymakers to include provisions specific to generative AI. Developers of such models are required to disclose detailed information about the data used for training, adhere to copyright laws, and label AI-generated content as manipulated when applicable. Enhanced scrutiny is reserved for the largest and most powerful AI models, given concerns about their potential for serious accidents and the proliferation of harmful biases.
Influence
Global influence: Europe’s pioneering AI regulations are expected to influence AI governance worldwide, setting a precedent for other governments grappling with similar challenges. While the United States and China are also advancing their AI agendas, Europe’s proactive approach underscores its commitment to shaping the future of AI regulation.
Implementation and enforcement: Following final approval, the AI Act is set to become law by mid-2024, with provisions phased in gradually. Each EU member state will establish its own AI watchdog for enforcement, while a central AI Office in Brussels will oversee compliance, investigation, and supervision. Violations of the AI Act could result in hefty fines, signaling the EU’s intent to uphold rigorous standards in AI governance.
As Europe moves closer to formalizing the AI Act, the world watches closely, anticipating its impact on the future development and regulation of artificial intelligence on a global scale.
Source: The Associated Press
