The European Parliament recently approved broad legislation governing artificial intelligence, including powerful systems like OpenChatGPT. The draft received approval from 523 EU parliamentarians, with 46 voting against. According to the EU Parliament, the 27 member states are anticipated to approve the law in April before it is published in the bloc's Official Journal in May or June. The AI Act focuses on higher-risk technology applications by the corporate and public sectors, imposing more robust requirements on providers, stricter transparency rules for the most influential models such as ChatGPT, and outright banning technologies deemed too harmful.
Senior European Union officials say the rules, first proposed in 2021, will protect citizens from the risks of a technology developing at breakneck speed while fostering innovation on the continent. EU chief Ursula von der Leyen hailed the vote as ushering in a "pioneering framework for innovative AI, with clear guardrails." "This will benefit Europe's fantastic pool of talents and set a blueprint for trustworthy AI worldwide," she said on X.
Global AI Race
Brussels has been racing to pass the new laws since OpenAI's Microsoft-backed ChatGPT debuted in late 2022, sparking a worldwide AI race. There was a surge of interest in generative AI after ChatGPT impressed the world with its human-like capabilities, which included digesting dense text, generating poems in seconds, and passing medical exams.
Other examples are DALL-E and Midjourney, which generate pictures, and others that create sounds based on simple input in daily English. However, the joy was quickly replaced by recognising the concerns, particularly that AI-generated audio and video deepfakes would turbocharge disinformation efforts. "Today is another historic day in our long journey to regulate artificial intelligence," said Brando Benifei, an Italian politician who pushed the legislation through parliament alongside Romanian MEP Dragos Tudorache.
"We managed to find that delicate balance between the interest to innovate and protect," Tudorache told journalists before the vote. Rules governing AI models such as ChatGPT will take effect 12 months after the law is passed, while most other provisions must be met within two years.
AI policing limits
The EU's "AI Act" regulations take a risk-based approach: the riskier the system, the more stringent the criteria, including outright prohibitions on the AI technologies that pose the most significant harm. For example, high-risk AI suppliers must assess risk and ensure that their products conform with the law before making them available to the general public.
Violations can result in fines ranging from 7.5 million to 35 million euros ($8.2 million to $38.2 million), depending on the type of infringement and the company's size. AI for predictive policing is strictly prohibited, as are systems that use biometric data to determine an individual's race, religion, or sexual orientation.
The guidelines also prohibit real-time facial recognition in public places, with certain exceptions for law enforcement; nonetheless, police must obtain judicial approval before deploying AI. Tudorache, a lawmaker, acknowledged that the law was heavily lobbied, but emphasised that they rejected the pressure. In a joint statement, organisations representing the European creative and cultural sectors welcomed the vote. Still, they asked the EU to ensure "these important rules are put into practice in a meaningful and effective way."