Brussels: The European Parliament has granted its final endorsement to a comprehensive set of regulations aimed at governing artificial intelligence (AI), which includes influential systems such as OpenAI’s ChatGPT. Garnering substantial support, the legislation was passed with 523 votes in favor and 46 against.
Timeline for Implementation:
The EU’s 27 member states are anticipated to ratify the law in April. Following this, the legislation is expected to be published in the EU’s Official Journal by May or June, as per the EU Parliament’s schedule.
Key Aspects of the AI Act:
The AI Act is primarily concerned with the regulation of high-risk AI applications within both the private and public sectors. It imposes stringent obligations on AI providers, enforces stricter transparency requirements for leading AI models like ChatGPT, and prohibits the use of AI tools deemed excessively hazardous.
EU Officials on the AI Act:
EU officials, who have been advocating for these regulations since their initial proposal in 2021, assert that the rules will shield citizens from the potential dangers of rapidly advancing AI technologies while simultaneously promoting innovation across Europe. Ursula von der Leyen, President of the European Commission, praised the legislative vote, describing it as the inception of a “pioneering framework for innovative AI, with clear guardrails.”
The Global AI Race:
The urgency to enact these new regulations has intensified following the debut of OpenAI’s Microsoft-supported ChatGPT in late 2022, which sparked a worldwide AI race. The generative AI’s human-like abilities, ranging from interpreting intricate texts to crafting poetry and passing medical examinations, have been met with widespread enthusiasm.
AI’s Double-Edged Sword:
While AI systems like DALL-E and Midjourney have demonstrated their prowess in image generation, and others in sound creation, there’s a growing awareness of the potential threats they pose, particularly the risk of AI-generated deepfakes fueling disinformation campaigns.
Balancing Innovation and Protection:
Italian MEP Brando Benifei and Romanian MEP Dragos Tudorache, who played pivotal roles in advancing the legislation through parliament, emphasized the delicate equilibrium achieved between fostering innovation and ensuring protection.
Enforcement and Compliance:
The provisions specific to AI models akin to ChatGPT will be effective 12 months post-officialization of the law. Other stipulations require compliance within two years. The AI Act adopts a risk-based approach, with more stringent regulations for systems considered higher-risk, including mandatory risk assessments and legal compliance for high-risk AI products prior to public availability.
Penalties for Non-Compliance:
Infringements of the AI Act could result in substantial fines for companies, ranging from €7.5 million to €35 million, based on the nature of the violation and the company’s size.
Prohibitions and Exceptions:
The legislation strictly forbids the use of AI for predictive policing and the analysis of biometric data to deduce characteristics like race, religion, or sexual orientation. It also bans real-time facial recognition in public spaces, with certain exceptions for law enforcement subject to judicial approval.
Lobbying and Resistance:
Despite being one of the most heavily lobbied pieces of legislation, MEP Tudorache affirmed that the parliament withstood the pressure.
Response from Cultural and Creative Sectors:
Representatives of the European creative and cultural sectors have expressed their support for the legislation. However, they urge the EU to ensure the effective and meaningful implementation of these critical regulations.
