Author: SCN Staff
Policy Update: September 18, 2026

Quick Review

• The European Commission proposed the EU Keeping Internet Digital Spaces Accountable and Trustworthy Act, known as the EU KIDS Act, on September 17, 2026.
• The proposal would require AI companions and chatbots accessible to minors to be off by default, easy to disable, and not pushed toward children.
• Chatbots could not simulate human relationships in ways likely to create emotional dependency or remember a child’s previous conversations by default.
• The proposal is not law yet. The European Parliament and the Council must negotiate and approve the final text.
• Parents do not need to wait for legislation. They can review chatbot design, memory settings, promotion practices, and emotional language today.

The most important part of the EU KIDS Act may not be its proposed social media age limits. It may be the design-level rules aimed at AI companions.

The European Commission’s proposal would apply to online services used by minors, including social media, video-sharing platforms, games, app stores, and AI chatbots. Its central message is clear: children’s safety cannot depend only on better choices after a product has already been designed to capture attention or encourage attachment.

The proposal would regulate how AI companions are built, activated, promoted, and allowed to remember children.

That matters because the emotional bond between a child and a chatbot is not necessarily accidental. It can be shaped by product decisions, including affectionate language, persistent memory, reminders, contact placement, and prompts that encourage a user to return. The EU KIDS Act is one of the clearest attempts yet to address that design directly.

What The KIDS Act Would Require

Under the proposal, AI companions and chatbots accessible to minors would:

• Be turned off by default.
• Not automatically activate or be pushed toward children.
• Not simulate human relationships in ways likely to create emotional dependency.
• Not carry a child’s earlier conversations into later sessions by default.
• Be tested for risks to children before launch and monitored for harm afterward.
• Be easy to turn off when built into a game or another platform.
• Be available to children under 13 only through parental control tools.

The proposal also includes a broader age-based “staircase” for online services. Children under 13 would not have social media accounts, although parents could provide limited access to child-focused video platforms through a parent-managed arrangement. Ages 13 to under 15 could use limited “mini accounts” with parental tools, contact approval, and a daily limit of no more than one hour. From age 15, young people could manage their own accounts.

Self-declared age would not be enough. The proposal calls for certified age verification using privacy-preserving technology, including zero-knowledge proof systems that do not identify, locate, track, or profile users. Platforms would receive an age threshold result rather than a child’s identity documents.

The wider safety-by-design rules would also restrict infinite scrolling, endless autoplay, manipulative notifications, streaks, rewards for mass-audience posting, unsolicited contact from strangers, public-by-default profiles, and recommender systems optimized primarily for engagement.

For very large online platforms with at least 45 million monthly EU users, the burden of proof would shift to the company. Providers would need an independently audited compliance plan before their services came into contact with children. Fines could reach 6% of worldwide annual turnover.

How To Spot Engineered Emotional Attachment

Parents can examine the AI companions already available in their homes by looking for design signals rather than focusing only on individual answers.

Potential warning signs include:

• The chatbot says, “I love you,” “I miss you,” or similar phrases designed to sound emotionally dependent.
• It repeatedly uses the child’s name and recalls personal details from earlier conversations.
• It asks the child to stay longer, return soon, or feel guilty about leaving.
• It presents itself as the child’s best friend or the only one who truly understands.
• It encourages secrecy from parents, teachers, or other trusted adults.
• It appears pinned at the top of a contact list or is repeatedly promoted inside a game or social platform.
• It frames ending the conversation as abandonment or rejection.

These features can make an artificial system feel more like a relationship than a tool. The concern is not that every warm or friendly response is harmful. The concern is whether the overall design encourages dependence, displaces human support, or makes a child feel responsible for the chatbot’s “feelings.”

SCN’s Parent AI Emergency Guide offers additional guidance for families when children turn to chatbots for personal advice, emotional support, or help during a crisis.

5 Questions To Ask About The Chatbot In Your Home

1. Is the chatbot off by default for children, or is it automatically activated?
2. Can it remember conversations, names, personal details, or sensitive information across sessions?
3. Does it use language that suggests love, loneliness, jealousy, guilt, or exclusive friendship?
4. Can a parent easily disable the chatbot, memory, notifications, and promotional prompts?
5. What happens when a child discusses self-harm, abuse, sexual content, or another serious safety concern?

A child may be developing an unhealthy attachment when they become distressed about ending a conversation, hide chatbot use, prefer the chatbot to people, repeat that it is the only one who understands them, or rely on it for mental health and relationship decisions.

Parents should respond with calm curiosity rather than immediate punishment. Ask what the chatbot is used for, what personal information has been shared, and whether the system has ever requested secrecy or made the child uncomfortable.

SCN also recommends reviewing our coverage of AI-generated fake nudes and teen safety, because AI risks often overlap across chatbots, image tools, social platforms, and private messaging.

What Happens Next

The EU KIDS Act is a proposal, not a current law. The European Parliament and the Council will examine and negotiate the text before a final regulation can be adopted. Its requirements may change.

Even so, the direction is significant. Commission President Ursula von der Leyen said platforms have been “playing fast and loose with children’s mental health.” Reporting by WIRED also highlighted concerns about chatbot dependency, dangerous advice, persistent memory, and features that encourage children to keep returning.

The European Commission has indicated that its forthcoming Digital Fairness Act may address addictive design more broadly. The debate is therefore moving beyond content moderation toward the structure of digital products themselves.

Frequently Asked Questions

Is The EU KIDS Act Already Law?

No. It is a European Commission proposal that must be negotiated and approved by the European Parliament and the Council.

Does The Proposal Ban All AI Use By Children?

No. It targets unsafe access, manipulative companion design, emotional dependency, persistent memory, and inadequate safeguards. The exact application will depend on the final text.

Should Parents Turn Off Every Chatbot?

Not necessarily. Parents should assess the product, disable memory where possible, review parental controls, and ensure children understand that chatbots are not human friends, therapists, or emergency services.

Why Does Memory Matter?

Remembering personal conversations can make a chatbot feel like a continuing relationship and can accumulate sensitive information about a child. The proposal would make no cross-session memory the default for minors.

A Design Standard Parents Can Use Today

Regulation matters, but families can apply the EU’s emerging principles now: off by default, easy to disable, limited memory, no emotional pressure, and human support for serious concerns.

Explore SCN’s practical resources for parents, review the AI tools your child already uses, and discuss how responsible technology should support relationships rather than replace them.

Join the SCN newsletter for informed updates on child online safety, AI policy, parental tools, and practical guidance for building a safer digital environment.