As machine learning and artificial intelligence technologies shape our current era, it is important to note that there is a dire need for regulatory intervention in the field. It isn’t surprising that although the field of AI is supposed to “do good” for the advancements of human problems, there is a downside to all technological innovations. While several AI systems pose minimal risk at present, they need to be assessed. Several regulations/acts have been drafted and released across the world to target the potential risks of the technology.
While the EU AI act was one of the very first regulations to be drafted, the rest of the world, including China and the US, have not been too far behind. The EU AI act is targeted at managing risks with systems related to law enforcement, education and immigration with the regulation aiming to assess them before being put into the market for use. If one were to look at the regulatory impact, companies that fail to comply will be fined 6% of their annual turnover. Nearly 20-25% of the global AI market comes in from the EU region, so the EU AI Act cannot be ignored.
There are many reasons regulatory intervention helps AI systems and their impact on the public. First and foremost, regulation is needed to protect the privacy of users and their data. This is especially true for data-heavy algorithms, with potential to remove bias and discrimination. Regulatory frameworks need to enforce mechanisms whereby bias is not purposefully injected and ensure that naturally occurring biases are removed. Then there are human rights and safety concerns, where deep fakes can be avoided or misinformation is not spread. Finally, it is imperative that AI-led development does not monopolize and favor certain human populations while overlooking the majority whose data isn’t used to train algorithms.
As the saga of regulatory intervention begins, governments around the world will need to collaborate and establish broad regulatory frameworks. There is also a dire need for education and knowledge sharing on developing machine learning techniques and algorithms. In the current context, with respect to the EU AI act, no foundational LLM (Large Language Model) complies with its clauses. China’s regulation focuses on control of content and is not too concentrated on the risks. These frameworks need to be inclusive and adaptive, as well as be updated from time to time.
Furthermore, we know that regulation, bills and laws often fail to adapt to changing technology. Machine learning and artificial intelligence is a rapidly evolving field, with new techniques and applications emerging faster than you would imagine. New challenges, risks and opportunities continuously emerge, and we need to remain agile/flexible enough to deal with them. Keeping up with the advancements and regulating cutting-edge technologies can be challenging for governing bodies. This needs to be factored in as regulatory frameworks for AI continue to evolve.
Editor’s note: For further insights on this topic, read Shini Menon’s recent Journal article, The Potential Impact of New AI Regulations, ISACA Journal, volume 1, 2024.
