AI companies are facing growing pressure to strengthen protections for children and teenagers as the use of AI tools expands across education, search and everyday communication, raising concerns about potential risks for younger users.

Tom Siegel, a former vice president of trust and safety at Google, warned that the technology industry risks repeating mistakes made by social media companies in dealing with the impact of digital products on children. He said the interactive nature of AI could create additional challenges as younger users increasingly engage with these systems.

Siegel has joined Common Sense Media as the first executive director of its Youth AI Safety Institute, a research initiative focused on studying the impact of AI on children and developing independent standards to assess potential risks. The institute is backed by organizations including Anthropic and the OpenAI Foundation.

Siegel has identified several potential risks, including suicide and psychosis, as well as what is known as “cognitive offloading,” in which users rely on AI for answers instead of developing their own critical-thinking skills.

He has called for stronger age-verification systems and controls that would prevent children from accessing services that do not provide adequate safeguards.

Google says it already offers protections for younger users, including tools that allow parents to disable their children’s access to search. The company also highlights the potential of AI-powered search to support learning and exploration.

OpenAI said it uses account and behavioral signals, along with users’ self-reported ages, to help identify accounts that may belong to underage users. The company also said it is testing additional methods to prevent users from bypassing its safeguards. Anthropic did not immediately respond to a request for comment.

The debate is increasingly focused on whether existing safeguards are strong enough to keep pace with the rapid expansion of AI tools and their growing user base.

Siegel has proposed independent testing of AI products, similar to safety assessments used in the automotive industry. Under such a system, outside organizations could evaluate publicly available AI tools against common standards designed to identify risks to children.

Such standards could eventually become an important factor in the AI market, particularly for parents, schools, regulators and institutional customers. At the same time, companies that expand their products without sufficient investment in safety measures could face greater legal and regulatory risks.

The debate comes as the AI industry works to establish safeguards for younger users while the technology continues to expand into everyday life. Cooperation between companies and independent safety initiatives, along with decisions by governments on whether voluntary standards should become mandatory, could shape how AI products are developed and used by children in the years ahead.