Why OpenAI is finally paying attention to teen users

Nearly 90% of teens already use ChatGPT weekly — for homework, research, or just getting their chaotic lives organized. That number is staggering, and it explains why OpenAI has decided to stop treating teenagers as an afterthought.

The company’s latest update isn’t just another content filter slapped on top. It’s a full rethink of how the chatbot behaves for younger users, wrapped around a simple argument: keeping teens away from AI entirely would leave them unprepared for a technology that’s becoming as fundamental as the internet itself.

So what actually changes? Let’s break it down.

Age prediction: the invisible safety net

ChatGPT now uses age prediction to automatically apply a more age-appropriate experience whenever it estimates a user is under 18. If the system isn’t sure, it defaults to the safer teen experience. That means stronger safeguards against graphic violence, self-harm content, unhealthy body image, and those risky viral challenges that sweep through school hallways.

It’s not a perfect system — no AI age detection is — but the default-to-safe approach is a smart move. Teens who lie about their age will still hit the stricter guardrails more often than not.

Break reminders: because teens forget to sleep

Anyone who’s watched a teenager disappear into a screen for four hours knows the struggle. OpenAI is adding more frequent break reminders during long sessions, gently nudging users to step away. It’s a small feature, but for parents fighting the endless battle against bedtime, it’s a welcome one.

Parental controls that actually do something

The expanded parental controls are where this update gets real. Parents can now:

• Set Quiet Hours to block usage during specific times
• Disable Voice Mode entirely
• Manage access to image generation
• Receive notifications in high-risk situations, including signs of potential self-harm

That last one is significant. Real-time alerts about self-harm signals could genuinely help parents intervene early. It’s the kind of feature that moves beyond convenience into genuine protection.

Study Mode: learning instead of cheating

Here’s the part educators will appreciate. OpenAI has rolled out Study Mode, developed alongside teachers and learning experts. Instead of handing over answers, it walks students through problems step by step. The goal is to make ChatGPT a tutor, not a shortcut.

Parents can now turn Study Mode on by default for their teen’s account directly through parental controls. OpenAI has also added new starter prompts built for schoolwork, plus interactive math and science tools that now reach 18 million weekly users across more than 250 topics.

That’s a lot of teens doing actual learning — or at least pretending to while the app guides them through the process.

What this means for parents and teens

The bigger picture here is that OpenAI is acknowledging what most parents already know: banning AI outright doesn’t work. Teens will find ways around restrictions, whether it’s using a friend’s account or a different tool entirely. Building safety features into the experience is a more realistic approach than trying to keep kids away from the technology altogether.

For parents, the new controls offer a middle ground between total surveillance and blind trust. Quiet Hours, voice mode limits, and self-harm alerts give you concrete ways to shape how your teen uses ChatGPT without hovering over their shoulder.

For teens, the trade-off is clear: you get access to a powerful learning tool, but you lose the ability to use it as a 24/7 homework machine with zero oversight. That seems like a fair deal.

As AI becomes more embedded in daily life, expect other platforms to follow OpenAI’s lead. The question isn’t whether teens will use AI — they already do. The question is how we make that use safe, productive, and honest. This update is a solid step in that direction.

If you’re a parent setting up your teen’s account, start by exploring the parental controls in the settings menu. Turn on Study Mode, set your Quiet Hours, and have a conversation about what the self-harm alerts mean. The tools are there — it’s up to you to use them.

A Search Engine That Knows You’re Busy at 7 PM

Ask Google Search for a dinner recommendation, and it might now check your calendar first. That’s the quiet but significant shift the company just announced: AI Mode’s Personal Intelligence can now connect directly to Google Calendar. Not just to read your schedule — but to create events on your behalf.

The feature is rolling out to users in the United States starting now, with a wider international release planned later. It’s a small step in the interface, but a giant leap in what Search is becoming.

Calendar Adds Action to Personal Intelligence

Until now, Personal Intelligence mostly pulled context from Gmail and Google Photos. Those connections made responses more relevant. Calendar changes the game in a different way: it’s the first connected Google app that doesn’t just inform. It acts.

Robby Stein, Google’s Vice President of Product for Search, explained that AI Mode can now understand what’s already on your calendar before answering. Suppose you ask for restaurant ideas. Instead of a generic list, AI Mode might notice you have an evening meeting and recommend a spot that fits your actual window of free time.

And if someone sends you an invitation or mentions a meeting in a message, AI Mode can create the corresponding calendar entry without you ever opening the Calendar app.

From Gmail and Photos to Time Awareness

Gmail and Photos provided context — your emails, your pictures. Calendar introduces something entirely new: time awareness. Two people asking the exact same question could now get completely different answers, simply because their schedules differ.

Google first previewed Calendar integration at Google I/O 2026, though it didn’t announce a timeline then. Now it’s live, joining Gmail and Google Photos in the broader Personal Intelligence ecosystem, which has grown from AI Pro subscribers to nearly 200 countries and 98 languages.

What This Means for Search Results

For decades, the same query produced essentially the same results for everyone. That assumption is crumbling.

Research from SEO firm iPullRank found that connecting Gmail to Personal Intelligence changed which brands appeared in AI-generated responses, even with identical prompts across different accounts. Calendar adds another layer of personalisation that goes beyond preferences — it’s about availability, commitments, and timing.

This raises a fundamental question: if Search answers tailor themselves to your schedule, how do you verify what’s accurate? You’re no longer looking at one definitive result page. You’re looking at an answer shaped by dozens of personal signals that exist only inside your own Google account.

The Road to a Proactive Assistant

The logical next step is obvious. If Google connects apps like Keep, Tasks, Maps, Docs, or even third-party productivity platforms, AI Mode could shift from reactive to proactive. Instead of waiting for you to ask, it might anticipate what you need next.

That’s a very different kind of search experience.

It also makes Google’s AI search updates harder to evaluate. Instead of verifying one result, you’re trusting a system that reasons with your personal data. The trade-off is convenience for transparency.

Is This the Future of Search?

Google’s move suggests the future of Search isn’t about finding the same answer as everyone else. It’s about finding the answer that makes the most sense for you — right now, given your schedule, your commitments, and your time.

For users, that’s genuinely useful. For anyone who relies on predictable search results — marketers, publishers, researchers — it’s a shift that demands attention.

Whether you see it as a helpful assistant or a privacy concern, one thing is clear: Search is no longer just a tool for finding information. It’s becoming a personal planner that knows when you’re free.

The $110 million bet on cells, not just algorithms

When GSK expanded its partnership with London-based Relation Therapeutics in a deal worth up to $110 million, the pharma giant wasn’t just buying another AI platform. It was paying for something far scarcer: high-quality biological data.

The collaboration centers on generating large-scale datasets that measure how human cells respond to genetic changes and drug interventions. That data will train AI models to spot potential drug targets, including those within Relation’s MORGAN platform.

It’s a telling shift. For years, the buzz in drug discovery has been about smarter algorithms. Now, the bottleneck is moving to the fuel that powers them.

What Relation Therapeutics actually does

Relation calls its approach Lab-in-the-Loop. It’s a cycle: run lab experiments, feed the results into computational models, use those models to design the next experiment. The company handles tissue profiling, single-cell and spatial transcriptomics, sequencing, and target validation. Machine learning sits at the center, guiding target identification and experimental design.

Its perturbation experiments are particularly interesting. These measure how genetic tweaks alter cellular characteristics tied to disease. The results get analyzed alongside genetic and patient-derived data, creating a richer picture than any single dataset could offer.

This isn’t theoretical. Relation’s Osteomics project is a proprietary functional single-cell bone atlas, built from patient samples and combining single-cell and spatial omics with imaging, genomics, proteomics, and clinical phenotypes. It’s already being used to investigate osteoporosis biology and identify patient subgroups, with hospitals in the UK and Australia involved.

Why bigger datasets don’t automatically mean better AI

Here’s where things get counterintuitive. You might assume more data always helps. A June 2025 study in Nature Methods suggests otherwise.

Researchers trained 400 single-cell foundation models on a corpus of 22.2 million cells, evaluating them across 6,400 experiments. The result? Models hit performance plateaus after training on only a fraction of the available data. Unlike large language models, these systems didn’t show clear scaling laws where more data consistently led to better results.

The study concluded that model capacity, dataset size, and compute need to be balanced—not just cranked up together. Adding more biological data didn’t reliably produce better models.

A separate 2025 study in Genome Biology evaluated Geneformer and scGPT, two well-known single-cell foundation models. Neither consistently outperformed simpler approaches. The researchers also flagged batch effects and warned against assuming larger pretrained models automatically yield better biological representations.

Quality control is the real challenge

The problem isn’t just volume. It’s consistency. A 2025 review in Experimental & Molecular Medicine noted that public repositories like CZ CELLxGENE, the Human Cell Atlas, and NCBI Gene Expression Omnibus offer vast amounts of single-cell data—CZ CELLxGENE alone has over 100 million standardized cells.

But that data comes from hundreds of labs, each with its own sampling methods, sequencing protocols, and processing pipelines. Technical noise and artifacts are everywhere. Dataset overlap is another headache: the same cells can appear in multiple resources, giving them outsized influence during training and creating data-leakage risks when training and test sets overlap.

The review’s conclusion was blunt: assembling a high-quality, non-redundant dataset matters just as much as model architecture.

Pharma’s new playbook: specialized data as a strategic asset

This is why companies like GSK are paying premium prices for proprietary data. A 2025 Nature Biotechnology analysis of AI-focused biopharma deals identified specialized dataset providers as a key trend, alongside larger upfront payments and new therapeutic modalities.

The analysis pointed to several examples:

• GSK’s separate $37.5 million agreement with Ochre Bio for human liver single-cell and perfused-organ data
• AstraZeneca and Pathos AI’s $200 million deal with Tempus in 2025, covering de-identified clinical, genomic, and imaging data from over 150,000 patients
• Relation’s own Osteomics atlas, built specifically for osteoporosis research

These deals reflect a simple reality: high-quality, disease-specific datasets are becoming a critical input for causal and generative machine-learning models. Public data has its place, but it’s not enough.

The data bottleneck won’t disappear anytime soon

A Nature research highlight on federated learning in pharma called limited access to suitable training data a major bottleneck for AI applications. Companies face restrictions on sharing proprietary information, and even when they want to collaborate, the data infrastructure often isn’t there.

That’s why AI-biopharma agreements take so many forms. Some focus on accessing AI platforms. Others cover joint development or data licensing. The GSK–Relation deal does both: it funds data generation and model development, with Relation producing human cellular datasets and using them to train target-identification models.

The message is clear. In AI drug discovery, the algorithm is no longer the differentiator. The data is. And the companies that control the best biological datasets—not the biggest ones—will likely lead the next wave of discoveries.

For more on how AI is reshaping pharma, check out how AI is shortening drug discovery timelines in China and AI foundation models in biomedicine.