![Story image](https://securitybrief.co.uk/uploads/story/2024/07/24/techday_9e5e12ad276313137abc.webp)
Experts weigh in on the complexities of the EU AI Act implementation
The European Union’s Artificial Intelligence Act (EU AI Act), set to come into force on 1 August 2024, is being hailed as a landmark regulation that aims to create a common regulatory and legal framework for AI within the European Union. With this development, the discourse around its implications and the challenges it presents to businesses and the tech community has intensified.
Dr Leslie Kanthan, CEO and Co-founder of TurinTech, views the Act as a significant but complex development for the AI industry. "The EU AI Act represents a crucial development for the AI industry, but it also introduces another layer of complexity for businesses to contend with, as they now must balance innovation and business growth with compliance. With it soon to be followed by initiatives like the UK AI Regulation Bill, compliance is becoming an integral part of any AI operation," Kanthan stated. He emphasised the need for businesses to thoroughly understand their AI code, ensuring each line adheres to the new legal standards. This, he believes, will not only maintain compliance but also result in optimal and efficient AI operations.
Sven Albrecht, Vice President of Engineering at Navenio, highlighted the dynamic nature of AI technology and the challenges it brings to assessing risks. "Given that AI technology is rapidly evolving, it’s often hard to assess potential risks as they’ll only become apparent once new systems are deployed to users. Effective regulation must strike a delicate balance: It needs to be robust and flexible enough to address emerging threats without stifling innovation, ensuring that the EU is competitive in the global AI landscape," Albrecht remarked. He stressed that an agile and adaptive regulatory approach is essential to harness the benefits of AI while mitigating its risks, advocating for regulations that do not impose prohibitive compliance costs on smaller companies to avoid market monopolies by tech giants.
Ben Wells, a Data Scientist at Bayezian, addressed the EU AI Act's potential impact on innovation and the responsibility of AI model creators. “While regulation is welcome, the EU AI Act must strike the difficult balance between effectiveness without stifling innovation. Debate continues over who should be regulated, as some creators argue they shouldn’t bear the full weight of responsibility. Unfortunately for them, it makes sense for models to be regulated at the base level; otherwise, misuse will always occur if the model is capable of it,” Wells commented. He argued that the compliance timeline, ranging from 24 to 36 months, might be too slow given the fast-paced evolution of AI technology, suggesting that updated rules might be necessary even within the compliance window.
Sebastian Gierlinger, Vice President of Engineering at Storyblok, discussed the ramifications of the AI Act on content creation and copyright policy. "The EU AI Act has established a framework of best practices to govern the use of AI systems. Experts have called for an international agreement on using copyright-protected content to train AI models, such as watermarking AI-generated content with significant human input, intellectual property (IP) protection of copyright-generated content, an enforced liability for infringements when creating content with Gen AI, and a remuneration scheme for rightsholders," Gierlinger elaborated.
While acknowledging the challenges generative AI poses to copyright law, Gierlinger noted that the Act includes transparency requirements for publicly disclosing the use of copyrighted training data. Some exceptions, such as for research institutions, are noted, but these do not apply to commercial AI systems. The AI Act aims to balance copyright protection with promoting research and innovation and recognises the need for proportional compliance requirements for startups and SMEs.
As the AI Act approaches implementation, Gierlinger advices companies to develop comprehensive AI policies ensuring ethical, legal, and effective use of AI technologies. The framework set by the Act is expected to provide much-needed accountability and transparency in AI operations across the EU.