Artificial intelligence is moving faster than any consumer technology we have seen before. New models launch weekly, products iterate in months rather than years, and new products and agents are popping up everywhere you look. While this pace can feel exhilarating, it’s also leading to models and products being launched haphazardly with safety and trust not built in as first-order design constraints but as afterthoughts.

Children are not just smaller adults navigating digital spaces. They are developing cognitively, emotionally, and socially in environments increasingly shaped by algorithmic decisions. As AI systems become more autonomous, more personalized, and more persuasive, the stakes of getting child safety wrong rise dramatically. We saw how social media built fast and broke things and addressed child safety as an afterthought leading to a youth mental health epidemic and an “Anxious Generation” of kids who are completely addicted to their phones. The leaders in AI are largely making all the same mistakes as those social media companies - optimizing for speed and engagement without being proactive about important and obvious child safety measures.

As we look toward 2026, AI builders, consumer tech companies, and regulators face a defining moment. This will be a pivotal year to see whether the industry will choose to design systems with children’s best interests at the core and/or whether the government will step up to create the necessary guardrails to keep kids safe. The decisions made this year will determine whether AI becomes a protective layer that quietly supports healthy childhood development, or an accelerant of existing harms.

AI builders and tech companies have responsibility to build products with trust and safety at the core. Meanwhile, the government has a key role to lay out AI guardrails, design requirements and enforcement mechanisms. Both groups should keep these trust and safety design principles in mind:

1) AI age estimation is not a replacement for strong parental controls

Ensuring kids are age verified correctly is a critically important part of protecting them from inappropriate content like porn to social media content and messaging older adults on the internet. Over the past decade, most technology companies simply asked the kids their age or birthdate which was insanely easy to circumvent. Obviously this is incredibly problematic and led to a flood of inappropriate content reaching kids and older - often dangerous - adults messaging kids which are both completely unacceptable. Research shows kids that watch age-inappropriate content (violence, mature content, etc) is linked to poor mental health outcomes.

One newer tool in the toolkit of age verification is AI age-estimation technology which will continue to grow more accurate, more ubiquitous, and more controversial. Advances in computer vision, behavioral analysis, and multimodal signals mean systems can increasingly infer age from how users interact, speak, or navigate digital environments—not just from what they claim on a signup form. This is particularly tricky because the companies with the data to do the best inference on age have a conflict of interest with financial motivations to verify ages older than the kids actual age to open up new ways of engagement and advertising. I saw a story last week that a mom posted about her son whose age she set to 9 on Roblox got “age-estimated” up to 13 which opened him up to a host of inappropriate games and strangers. And there was no way for her to manually adjust it back to his real age in the parental controls. When talking to a friend’s 10 year-old son last week, he told me he just asked an older friend to help him get through the age estimation so he could show up older.

As you can tell, AI age estimation might be better than a simple age-gate but it’s not good enough to fully replace great parental controls. AI needs to empower parents and loop them into key decisions about their kids. All kids accounts should be required to have a linked parent account which should be parent-verified - meaning the parent account has proven they are over 18 through a credit card transaction or ID verification.

2) Healthy engagement should be an explicit goal.

Everyone who uses social media or a video streaming service realizes the power of a great “For You” feed AI-personalized to your preferences. It can make discovery of new content fun and delightful. However, the never ending short-form videos personalized to you is a powerful hook for digital addiction and one reason teens spend north of 5 hours a day on social media. Research suggests that watching short-form videos in rapid succession on autoplay is, at a minimum, correlated with decreased attention spans, increased depression and anxiety, and worse time management.

Healthy engagement needs to be an explicit goal of tech companies. AI personalization needs to be balanced with non-addictive UX designs. I was really happy to see that YouTube recently announced it will let parents remove and set time limits on Shorts - their name for short-form videos. This likely came on the heels of tons of pressure from unhappy parents but regardless, was a step in the right direction. This is exactly the type of thoughtful UX design and parental controls that are needed to balance personalization with digital addiction. Others include landing pages more like Netflix or Hulu that makes you opt in to content vs. having it pushed at you on autoplay like a social media feed. I would also love to see built in design tools to help people manage usage - like asking how long they want to use the app when they first start and then providing reminders and support for them to get off of it. Another great UX idea to limit addiction is “Smart Notifications” that don’t ping me for every message or update but instead understand which messages I want to be pushed to me.

I believe if tech companies try to look the other way and continue to optimize engagement over what’s good for customers, they will continue to make money in the short-term and lose customers in the long-term. People are getting fed up with experiences that eat time and have diminishing value and trust. New entrants will emerge that offer joy and connection without all of the addiction and noise.

3) Parental controls need to get easier with AI agents.

All signs are pointing to big breakthroughs in agentic AI in 2026 - where AI agents can do more complex consumer experiences. While it will be great to have agents help book travel and schedule more meetings and send follow up emails, I’d also like to see this technology improve parent controls. When building my company, Sage Haven, a safer AI-moderated messaging app for kids, we interviewed 314 parents all around the U.S. It was so clear from these interviews that there were great intentions among the vast majority of parents to set up parental controls and keep their kids safe online. It was also clear they were completely confused, overwhelmed, and exhausted by the current systems. Parents are asked to configure dozens of settings across devices, platforms, and apps, each with its own logic and terminology. The result is predictable: many give up, leaving children exposed not because of neglect, but because of cognitive overload.

This creates an amazing opportunity for new AI agentic experiences where parents can simply speak their wishes and the AI agents can help them execute this well (and avoid the multilayers confusing screens and menus needed to do it manually today). Examples could be: “Ensure my kid only has two hours of screentime today but don’t worry about Spotify.” or “I’m okay if my kid plays Roblox but make sure they are only talking to their real friends and not strangers.” If executed thoughtfully, AI can be used for good and this is one area I’d love to see some growth!

4) Governance is needed where incentives are misaligned.

Because incentives of technology companies - like Roblox and social media apps - are not always aligned with what’s in the best interest of customers, governance is critical. Without this, you’ll continue to see age inappropriate content targeted to kids, parental controls as an afterthought, easily circumventable “solutions”, and AI age verification systems mysteriously age verifying kids older than their actual age.

There is a groundswell of parents speaking out against technology companies and demanding more online protections for kids which in turn is leading to broad, bipartisan support for child online safety measures.

I predict 2026 will be the year that several meaningful federal and state government protections are passed in the U.S.

I’m heartened by the wave of age minimums on social media sites enacted by various governments (Australia, Denmark, etc) and various U.S. states. These types of broad constraints make it much easier on parents to build collective action and reduce pressure on kids. Naturally, large AI companies with deep pockets are doing their best to limit regulation of any kind but I deeply believe government guardrails and thoughtful enforcement is critical. These include:

• Age minimums for things like AI chatbots and social media sites.
• Age verification requirements including robust auditing mechanisms to ensure age verification systems are accurate and part of robust parental controls systems
• Limits on access to phones during school hours
• Design requirements on app design and parental controls - including features that support moderate usage and the ability to override an AI age estimation and limit the types of content shown to kids (like Shorts)
• Meaningful liability for tech companies who violate these new laws

2026 is a moment of choice

2026 is not just another year of iteration—it is a moment of choice. The systems being designed now will shape how an entire generation grows up with AI. Builders can choose to embed trust, safety, and child-centered design into the foundations of their products, regulators can choose to set clear and enforceable guardrails, and together they can realign incentives toward long-term wellbeing over short-term growth. If we get this right, AI can become a quiet force for protection, creativity, and healthy development rather than a source of addiction and harm. The window to act is still open—but it is closing fast. What we decide in 2026 will define not only the future of AI, but the future of childhood itself.