The EU AI Act represents a crucial step towards responsible AI development, deployment, and use of AI in the European Union. However, Lamprini Gyftokosta, Director of Artificial Intelligence and Human Rights at Homo Digitalis, raises serious questions about its effectiveness and implementation.
Ultimately, Gytfokosta believes that the EU AI Act is necessary but flawed. I recently spoke with her to find out why.
A Market-Driven Approach with Unclear Goals
According to Gyftokosta, the Act’s purpose extends beyond simply regulating AI. “What the Act is really trying to do,” she said, “is improve the functioning of the EU’s internal AI market. It aims to enable the free movement of AI systems and models across member states to prevent regulatory fragmentation and create a unified market.”
However, while she acknowledges that the Act is necessary to bridge the gap between existing regulations and constant advancements in AI technology and respects its intention to protect fundamental digital rights, she questions whether it truly achieves these goals.
Enforcement: A House Built on Sand
For Gyftokosta, the EU AI Act has major enforceability issues. She argues that the level of cooperation between national authorities, government agencies, and the EU needed to enforce the Act is unprecedented, and, as such, is skeptical of its feasibility.
“The envisioned AI governance structure relies on strong national-EU cooperation. While theoretically possible, I’ve never seen an example of truly effective collaboration of this nature,” she said.
Adding to this concern is the severe lack of resources allocated to enforcement bodies. Gyftokosta points out the stark contrast between the massive investments in AI development and the meager resources provided for oversight.
“While AI development receives significant financial support,” she said, “the resources allocated to regulatory bodies are woefully inadequate. Some enforcement agencies lack even the most basic facilities, like office spaces. How can authorities be expected to enforce the AI Act without funding?”
Interpretation Challenges: Lost in Translation?
Gyftokosta believes many of the Act’s key requirements are far too difficult to interpret, even for legal experts. She asserts that the staggered release of information, coupled with procurement deadlines that precede crucial guidelines, has fostered an environment of confusion and ambiguity.
“When organizations seek guidance, they often receive conflicting advice from different legal experts. Even when the commission itself offers guidance, it often fails to fully address these interpretive challenges.”
This interpretive complexity is compounded by what Gyftokosta sees as a widespread lack of understanding and awareness of the Act, particularly among younger generations. Her experiences reveal a significant knowledge gap.
“Students of all ages, from 10-18, across both public and private schools, lack basic knowledge of their rights, the implications of technology, and the potential consequences of AI,” she argued. “They’re simply not equipped with the knowledge they need, and that’s a problem.”
Digital Rights Protections: A Square Peg in a Round Hole
Gyftokosta further questions the Act’s suitability for one of its key goals: protecting fundamental digital rights. She argues that the Act has too broad a scope, meaning its protections for these rights are inadequate. Consequently, she believes that the protection of these rights should be left to existing legal frameworks, like GDPR, consumer protection laws and the Charter of Fundamental Rights.
“My discussions with relevant stakeholders suggest that the relative lack of strong provisions for fundamental rights in the AI Act will be covered by the Charter of Human Rights and other legal instruments,” she said.
Innovation is a Balancing Act
The News of the European Commission withdrawing the AI Liability Directive came at a point when the Trump administration rescinded Joe Biden’s 2023 executive order on AI, accusing it of stifling innovation.
While Gyftokosta has concerns about the EU AI Act, she does not share Trump’s view of regulation. “I don’t believe the EU AI Act hinders innovation,” she said. ” Safeguards are necessary to ensure developers and others understand the societal impact of AI technologies.”
The key takeaway is that while Gyftokosta believes that the EU AI Act is important, its success hinges on overcoming significant obstacles. Her critique drives home a problematic truth: regulating AI is incredibly complex, and we’re unlikely to perfect it any time soon.
Josh is a Content writer at Bora. He graduated with a degree in Journalism in 2021 and has a background in cybersecurity PR. He's written on a wide range of topics, from AI to Zero Trust, and is particularly interested in the impacts of cybersecurity on the wider economy.
The opinions expressed in this post belong to the individual contributors and do not necessarily reflect the views of Information Security Buzz.