Experts Call for Stronger AI Regulation in Education

These digital products are predominantly designed by adults for an adult world, often prioritizing capital accumulation over the well-being and empowerment of young users. This design philosophy deprives children of essential conditions for a flourishing childhood.

Children possess unique expertise simply by being children, a perspective that adults often lack. It is crucial to listen to their voices, as they consistently express desires for privacy, safety by design, and less distraction across various cultures. Incorporating their input can lead to digital products that genuinely support their development.

The capabilities of technology, such as AI's ability to write or personalize, do not automatically equate to educational value for children. While some technological advancements might align with educational benefits, many do not. Education should prioritize fostering critical thinking and genuine human interaction, rather than merely adopting technologies because they exist.

The intense focus on competing in the global AI race consumes significant political capacity, diverting attention and investment from fundamental societal questions. Essential needs, such as the requirements of schools and communities, are often overlooked in this pursuit. The current environment forces small-scale projects to conform to predefined, corporate-set terms, stifling independent innovation. This prioritization risks neglecting foundational elements necessary for societal well-being.

While these will not solve all AI-related issues or perfectly distinguish 'good' from 'bad' AI, they serve as essential boundaries in a rapidly evolving and potentially dangerous technological race. Bans on social media features, for example, illustrate early attempts to assert regulatory authority.

Factory acts for child labor provided an early regulatory framework, while the invention of the weekend emerged as a critical societal innovation from that era. This historical example demonstrates that innovation is not solely technical, but also involves significant social and regulatory adaptations.

President Trump's intervention with Anthropic regarding the UK's AI Security Institute exposed significant gaps in current sovereignty approaches. A proposed 'stack' for these countries would prioritize these rights over the current overemphasis on financing. Current strategies often fail to foster independent governance, leaving nations vulnerable to external influences.

These standards explicitly state that AI should not pretend to have an identity, personhood, or agency, nor should it speak in the first person. The consultation process focused on trade-offs rather than simple approval, ensuring a nuanced approach. A certification scheme for educational AI products is planned for implementation in the autumn to reinforce these standards.

The absence of a follow-up service for AI, once released into the wild, is compared to the inconceivable idea of a car without insurance or ongoing checks. This issue applies differently to fundamental AI models versus specific applications, but the general lack of accountability for deployed AI remains a significant problem. Without such oversight, the unintended consequences of AI are difficult to track and mitigate.

Human development is profoundly shaped by our environment, not merely by genetics, as illustrated by the 'devil's hole pupfish' analogy. The risk with AI and significant environmental changes is the unintended consequences on the conditions under which humans form and develop. This includes both 'information harms,' which affect what we know, and 'formation harms,' which impact who we become during childhood. Protecting these developmental conditions is crucial.

The current educational landscape often presents a dichotomy of total prohibition versus vague guidance when it comes to AI. A more effective approach is to treat AI adoption as a shared experiment, embracing a process of 'proof and mistake.' This method encourages inserting AI into the student curriculum, acknowledging that students will inevitably need to use these tools in their future careers. Engaging students in this experimental process can foster adaptability and critical understanding.

Specifically, banning features that allow chatbots to form emotional bonds or relationships with minors is critical, as such interactions are deemed inappropriate for children. Additionally, strict prohibitions against using student data for future model training and preventing chatbots from learning deep personal details about children are necessary to protect their privacy and well-being. These measures aim to create a safer digital environment for young users.

Progress has been noted in meaningful youth engagement, with a survey showing that the percentage of young people who reported not being asked about their education improved from 90% in 2020/2023 to 50% recently. While this marks significant improvement, the goal remains for 90% of youth to affirm they have been consulted. Achieving this requires the institutionalization of youth councils and ensuring that consultations are youth-led, transparent, and provide clear feedback mechanisms. This approach ensures young voices are genuinely heard and integrated into educational planning.

AI data centers consume significant amounts of power and water, often relying on 'dirty' energy sources due to default industry paths. Regulation must have 'teeth' to shift incentives towards sustainable practices and prevent widespread environmental damage. This is a crucial area where regulation can make a substantial difference, ensuring that technological advancement does not come at an unsustainable ecological cost.

With 90% of employers expected to adopt AI and two-fifths of worker skills needing to shift within five years, there is a significant risk in treating higher education solely as an engine for labor market preparedness. Its purposes extend beyond employment preparation to fostering critical thinking, civic engagement, and holistic development. Over-prioritizing job training could diminish the broader societal contributions of universities.

By 2030, 39% of core skills are expected to change, reflecting the rapid evolution of industry demands. Furthermore, the region's historic demographic dividend is narrowing, with seniors projected to outnumber children by 2055. This demographic transition means that higher education institutions can no longer solely rely on traditional 18 to 20-year-old cohorts. Universities must adapt to become engines for lifelong productivity, catering to a broader age range and continuously upskilling the workforce.

This failure has created a widening gap between university output and the demands of modern careers, with 50% of US graduates showing no improvement in critical thinking. Universities often reward deep analytical work, which is insufficient for careers increasingly requiring the ability to synthesize information from diverse sources. This systemic oversight leaves graduates unprepared for the complex challenges of the contemporary workforce.

Universities face a crucial call to coexist with communities rather than engaging in extractive research practices, particularly when developing AI policies. Researchers often enter communities merely to extract information, leaving local populations excluded from policy-making processes. It is essential for universities to act as partners, not just external observers, fostering collaborative relationships. This shift ensures that AI policies are developed with community input and benefit, rather than exploiting, local populations.

These shifts include the integration of advanced computational social sciences to study social phenomena at scale, requiring new mechanisms for research cooperation that move beyond rigid models. Beijing Normal University, for example, is conducting intelligent social experiments grounded in real educational practice. The overarching focus is on supporting SDG4 and fostering an inclusive, equitable ecosystem through these advancements.

Reflecting on historical resistance to technology in education, such as a 'moderately unrealistic' 1993 university manifesto to prohibit the internet, highlights key lessons. Students back then desired educators who could accompany them through profound social change, willing to investigate what new technology might enable, erode, and demand. This intergenerational mentorship is crucial, as failing to provide such guidance can lead to negative consequences for future generations. As an educator, neutrality is not a viable position in these transformative times.

Public education is increasingly reliant on private AI systems that governments often do not own or fully comprehend. Despite lacking full control, governments remain accountable for the impacts of these technologies. Crucially, the ecological impact and indigenous knowledge systems are largely ignored in current AI policy discussions. Advocates emphasize safeguarding public infrastructure before it becomes permanently tied to proprietary AI platforms.

Institutions must not compel students or educators to use AI, especially when no meaningful alternatives exist. Conscientious objection to AI in education must be possible without stigma or disadvantage, ensuring individual autonomy. This principle safeguards against mandatory adoption of systems that may raise ethical, privacy, or pedagogical concerns. Providing meaningful alternatives respects diverse learning approaches and protects educators' professional judgment, promoting a more ethical and equitable integration of technology.

This change marks a new annual rhythm for the event, which has traditionally been held in September. The shift aims to establish a new dynamic for the conference, influencing its planning and thematic focus. This adjustment could lead to different participation patterns and topical discussions for the international gathering.