Europe Move Toward Regulating Artificial Intelligence
The European Parliament has recently passed a “proposed law,” the EU AI Act, aimed at creating uniform regulations for artificial intelligence throughout Europe.
The EU AI Act, a proposed law that aims to unify AI regulations across Europe, recently achieved a crucial milestone, with the European Parliament voting overwhelmingly in its favor.
This Act represents the most comprehensive effort globally to regulate AI and is designed to address a host of potential AI risks, ranging from privacy invasions and propaganda to biases, copyright infringements, and the generation of illegal content.
Many aspects of the Act remain abstract, and the road ahead involves extensive negotiations among European agencies and member states.
The European Parliament recently voted overwhelmingly in favor of the EU AI Act, a proposed law that would standardize AI regulations across Europe. This “negotiating position” is a critical milestone on the road to a unified European approach to regulating Artificial Intelligence tools.
The proposed EU AI Act has sparked, and will continue to spark lively debates: is it an overdue move to rein in uncontrolled technology giants, or an "innovation-hostile" effort that could relegate Europe to an "AI backwater"?
There are still many steps to go before the Act goes into effect. Many discussions and negotiations still to come among and between various European agencies, as well as all the European member states. Yet, its advancement represents the most comprehensive global effort to regulate AI to date, putting Europe at the forefront of AI legislation compared to its counterparts in the United States and China.
Guarding Against the Risks
The Act is designed to address a broad range of potential risks:
Privacy Invasion: The AI Act prohibits the use of AI tools for law enforcement, including facial recognition ("biometric surveillance"), predictive policing, and "emotion recognition."
Propaganda and Misinformation: In the social media era, AI's potential for crafting compelling ads and propaganda is a concern. The Act requires all AI-generated content to be explicitly labeled. Whether labelling content as “AI-generated” will address concerns about misinformation is another question.
Bias and Discrimination: There are worries that AI algorithms used for tasks such as job applicant screening or loan approvals may perpetuate discrimination and bias. While the issue is well known, it has also, so far, been very hard to address, not least because AI algorithms are “black boxes” and it is challenging to figure out how to root out improper bias.
Generation of Illegal Content: Generative AI's power to create text or images brings the risk of misuse, such as generating illegal content like child sexual material. The Act calls for AI companies to establish safeguards to prevent such misuse.
Copyright Infringement: AI tools like ChatGPT are trained on vast amounts of publicly available – and often copyrighted – material. The proposed regulations would require AI companies to disclose the training data for their AI tools.
Of note, too, is what is not in the Act, at least as of now. There does not appear to be, for example, any attempt to address the existential risks of AI, i.e., that AI may take on a life of its own. So far, the focus appears to be protecting against the risk that powerful large institutions, such as large corporations or governments, will use AI against individuals.
One Step in a Long Journey
What will happen next remains uncertain. The process for creating EU-wide rules is complicated, with many stakeholders. The “proposed law” is just the start.
At the moment, many of the Act's principles are still somewhat abstract, and the real test will be how all the various agencies and member nations negotiate what they mean in practice.
Issues related to AI tools sweeping up information about individuals and potential copyright infringements due to AI training on copyrighted data still need to be resolved. Finally, implementing safeguards against AI tools generating illegal content will be a complex endeavor.
The journey toward the implementation of the EU AI Act is just beginning, and we can expect a flurry of debates, refinements, and negotiations in the years to come. In other words, there are many fights still to come. Watch this space.