Pressure grows for AI regulation focused on children’s safety

As regulators around the world struggle to put the genie of social media back in its bottle and curb some of the harms it has caused, there are increasing calls to get ahead of the regulatory curve when it comes to AI.

The UN has just wrapped its first-ever Global Dialogue on AI Governance, and more than 100 organizations from across the globe have come together to call on governments to “move from commitments to action to protect children in the governance of AI.” According to a post from 5 Rights Foundation, which is coordinating the effort, the coalition argues in a joint statement that “current approaches too often address individual harms after they occur, rather than tackling the commercial incentives and governance failures that produce unsafe systems.”

In short, the argument is that children are being ignored in the development of national AI strategies and governance frameworks, even though they are among the most enthusiastic users of new technology. It’s not that the tools aren’t available; it’s that they’re not being applied with kids in mind.

“Two years after the adoption of the Global Digital Compact, there is no shortage of principles, commitments, or guidance,” the letter reads. “Organizations, experts, and child-rights advocates from across the world have helped build a clear framework for protecting children in the age of AI.”

“The challenge is no longer knowing what to do. It is doing it.”

Leanda Barrington-Leach, executive director of 5Rights Foundation, says this isn’t a case of adults forcing their will on kids. Kids, she says, are asking for change. “Children have given us a clear diagnosis of the problem. They aren’t asking us to block AI innovation, but it shouldn’t be a case of cleaning up the mess after harm has happened either. They’re asking us to change the incentives that produce unsafe systems in the first place.”

Avoiding social media’s regulatory mistakes

The comparison to social media matters beyond the policy debate. The most popular chatbots powered by large language models (LLMs) are OpenAI’s ChatGPT and Anthropic’s Claude – both products of the Silicon Valley culture that gave us Facebook and X. OpenAI is notoriously the brainchild of Sam Altman, and Anthropic was formed by a group of former OpenAI employees. The same business model that drives Meta’s platforms, in which the user and their data is the real product, underpins these companies’ vision. And they are fast accumulating the same kind of wealth and political influence as billionaires like Mark Zuckerberg – which will inevitably be used to fund legal efforts to push back against inconvenient regulation.

It is perhaps too grandiose to propose that the core problem is not with tech, but with the philosophy of those building it. Then again, little doubt remains that the titans of Silicon Valley designed and built their tech precisely to be addictive, with no regard for its effect on children’s lives.

Barrington-Leach notes that, “as long as companies are rewarded for speed, engagement and data extraction rather than safety, we’ll keep treating the symptoms while the disease becomes endemic.”

“Governments need to think carefully about the desired outcomes and beneficiaries of AI and change the rules of the market, so the AI race is geared to support rather than undermine the next generation. Respecting children’s rights must become a condition of doing business, not an optional extra.”

Recommendations points to importance of testing

Signed by prominent human rights organizations, academics and non-profits, the joint statement declares that “too often, commercial incentives have been allowed to outrun accountability. This must change.”

“We can no longer allow children to be collateral damage in the race to develop and deploy AI.”

To stop that from happening (again), the organizations offer 10 recommendations. At the top is a requirement for precertification. “Companies must demonstrate that systems affecting children respect their rights and are safe before they are brought to market,” the letter says. This points to an emergent truth in the digital ecosystem: the “wild west” era of the internet is long gone, and a promise of safety is no longer sufficient. In the present moment, testing, benchmarking and other third-party evaluations are now key to building established trust.

Subsequent recommendations urge governments to “ban manipulative design practices that exploit children’s vulnerabilities (notably for commercial purposes), jeopardize their agency, and foster unhealthy dependency” and hold companies to account when they do not measure up. Laws should demand transparency, close loopholes and offer redress. Digital surveillance should be curbed.

Finally, legislation should deliver effective regulatory oversight, ensuring businesses maintain high standards of fairness, transparency, accountability, and oversight across the AI lifecycle.

“This includes requiring businesses to implement technical standards, industry codes, and terms of service that adhere to the highest standards of privacy, safety, and ethics, and ensuring compliance with them.”

“The debate is not whether AI firms have a duty to protect children,” says the letter. “It is whether governments are prepared to act on their obligations and the commitments they have already made.”

Which ultimately makes accountability the core problem. Laws can be made, fines can be imposed, bans can be threatened, and so on. But little will change as long as massive tech companies believe that operating with impunity brings no meaningful consequence. The AI genie – which everyone from CEOs to governments would have us believe will transform the world – came from the same bottle as social media. It will not go gently back inside.

Article Topics

age verification  |  AI  |  chatbots  |  children  |  digital trust  |  legislation  |  regulation