<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[TeenCue Engineering]]></title><description><![CDATA[How we build TeenCue, parental controls for AI. Notes on Apple's Screen Time APIs, Safari and browser extensions, and privacy-first AI for families. Parents nev]]></description><link>https://teencue.hashnode.dev</link><generator>RSS for Node</generator><lastBuildDate>Sun, 22 Nov 2026 18:43:56 GMT</lastBuildDate><atom:link href="https://teencue.hashnode.dev/rss.xml" rel="self" type="application/rss+xml"/><language><![CDATA[en]]></language><ttl>60</ttl><item><title><![CDATA[How we tell parents what a teen's AI chat was about, without showing them the chat]]></title><description><![CDATA[How TeenCue rates a teen's AI chats by topic and concern level so parents know when to talk. Parents never see what was typed.]]></description><link>https://teencue.hashnode.dev/how-we-tell-parents-what-a-teen-s-ai-chat-was-about-without-showing-them-the-chat</link><guid isPermaLink="true">https://teencue.hashnode.dev/how-we-tell-parents-what-a-teen-s-ai-chat-was-about-without-showing-them-the-chat</guid><category><![CDATA[privacy]]></category><category><![CDATA[llm]]></category><category><![CDATA[AI]]></category><category><![CDATA[Security]]></category><dc:creator><![CDATA[Logan Bates]]></dc:creator><pubDate>Fri, 20 Nov 2026 08:50:33 GMT</pubDate><content:encoded><![CDATA[<p>Most parental controls for teens were built around one idea: let the parent see everything. Every text, every search, every message. That design breaks down with AI chatbots.</p>
<p>Teens talk to ChatGPT, Gemini, Claude and Character.AI about things they won't say out loud to anyone. Stress, friendships, their bodies, their sexuality, whether they're okay. If a parent can read those chats word for word, two things happen fast. The teen finds out, and the teen moves to an app the parent doesn't know about.</p>
<p>When we built TeenCue, we started from a different question: what is the least a parent needs to know to make a good decision? Our answer was the topic and how concerning it is. Not the words.</p>
<p>This post walks through how that works.</p>
<h2>The output: topic and concern level, nothing else</h2>
<p>For every message a teen sends on a supported AI chat service, TeenCue produces a small record:</p>
<ul>
<li><strong>Service</strong>: which AI it was (ChatGPT, Gemini, Claude, Character.AI, Copilot and so on)</li>
<li><strong>Category and topic</strong>: what the message was about, at a level like "school", "emotional support" or "relationships"</li>
<li><strong>Concern level</strong>: how likely it is that a parent should check in</li>
</ul>
<p>That record is what the parent sees in the dashboard, the iPhone app and the weekly Teen Brief email. Parents never see what was typed.</p>
<p>A typical brief reads something like: <em>TeenCue noticed 1 signal that may be worth a conversation. It was related to leaning on AI for emotional support.</em> That is enough for a parent to know it's time to talk, and it leaves the teen's words with the teen.</p>
<h2>Rating the message</h2>
<p>To produce that record, the message has to be read by something. We use an AI rating model from Anthropic, the company behind Claude.</p>
<p>The flow is short:</p>
<ol>
<li>The teen sends a message to an AI chat service in their browser or in Safari on their iPhone.</li>
<li>TeenCue sends that message securely to the rating model with a prompt that asks for one thing: a structured rating (category, topic, concern level), not a summary and not a quote.</li>
<li>TeenCue stores the rating and discards the message text.</li>
</ol>
<p>Asking for a structured rating instead of free text matters. A summary is just the chat in shorter form. A topic plus a level is something a parent can act on without learning what their kid actually said.</p>
<h2>What we keep and what we don't</h2>
<p>We keep:</p>
<ul>
<li>The rating (service, category, topic, concern level)</li>
<li>Counts and timing, so a parent can see things like "24 AI chats across 3 services this week"</li>
</ul>
<p>We don't keep:</p>
<ul>
<li>The text the teen typed</li>
<li>The AI's replies</li>
<li>Other websites or browsing history</li>
</ul>
<p>We're also upfront about the one place text does exist briefly. Anthropic, the rating provider, deletes the message within 30 days, except for chats its own systems flag for safety, which it can keep for up to 2 years. That is Anthropic's retention policy, and we'd rather say it plainly than hide it behind "never stored."</p>
<h2>Ratings can be wrong</h2>
<p>A model rating a single message without context will sometimes get it wrong. "I want to die" means something very different after a math test than it does at 2am.</p>
<p>So TeenCue treats a rating as a reason to talk, not a verdict. The product language is deliberately soft ("may be worth a look") and the parent view shows patterns across a week, not a single alarming line. The goal is a calm conversation, not a confrontation built on one flagged message.</p>
<h2>Designed to be used openly</h2>
<p>TeenCue does nothing until it's paired with a parent account using a short code, and we tell parents to let their teen know it's installed. That's not just ethics. Monitoring that relies on secrecy fails the moment it's discovered, and teens always discover it. Monitoring a teen knows about, and that doesn't expose their words, is something they can live with.</p>
<h2>Where it runs</h2>
<ul>
<li><strong>iPhone</strong>: Apple's Screen Time APIs plus a Safari extension, no VPN or MDM. We wrote about that in <a href="https://teencue.hashnode.dev/how-we-keep-ai-chatbots-inside-safari-on-a-teen-s-iphone-without-a-vpn-or-mdm">How we keep AI chatbots inside Safari on a teen's iPhone</a>.</li>
<li><strong>Computers</strong>: an extension for Chrome, Edge, Firefox and Brave.</li>
</ul>
<p>If you're a parent, you can try it free for a week at <a href="https://teencue.com">teencue.com</a>. If you're an engineer working on privacy-preserving safety tools, we'd like to hear how you'd approach this differently.</p>
]]></content:encoded></item><item><title><![CDATA[How we keep AI chatbots inside Safari on a teen's iPhone (without a VPN or MDM)]]></title><description><![CDATA[How TeenCue uses Apple's Screen Time APIs and a Safari extension to flag AI chats for parents, with no VPN or MDM. Parents never see what was typed.]]></description><link>https://teencue.hashnode.dev/how-we-keep-ai-chatbots-inside-safari-on-a-teen-s-iphone-without-a-vpn-or-mdm</link><guid isPermaLink="true">https://teencue.hashnode.dev/how-we-keep-ai-chatbots-inside-safari-on-a-teen-s-iphone-without-a-vpn-or-mdm</guid><category><![CDATA[iOS]]></category><category><![CDATA[Swift]]></category><category><![CDATA[AI]]></category><category><![CDATA[privacy]]></category><dc:creator><![CDATA[Logan Bates]]></dc:creator><pubDate>Fri, 20 Nov 2026 07:23:06 GMT</pubDate><content:encoded><![CDATA[<p>Most teens now talk to AI chatbots every day. Pew Research found in early 2026 that about 64% of US teens use them, and roughly 3 in 10 use them daily. Only about half of parents think their teen uses them at all.</p>
<p>We're building TeenCue, parental controls for AI. The idea is simple: let a parent know when one of their kid's AI chats may need a closer look, by topic and concern level, without ever showing the parent what was typed.</p>
<p>On a computer that's a browser extension problem. On an iPhone it's much harder, and this post is about how we approached it.</p>
<h2>The problem: AI lives in apps you can't see into</h2>
<p>On iOS, third-party apps are sandboxed. There is no supported way for one app to read what happens inside another app, and that's a good thing. It also means a parental controls app can't see what a teen types into the ChatGPT, Gemini or Character AI apps.</p>
<p>The usual workarounds are heavy:</p>
<ul>
<li><strong>VPNs</strong> that route all traffic through a server</li>
<li><strong>Configuration profiles or MDM</strong> that take over the device</li>
<li><strong>Reading everything</strong>, which turns parental controls into surveillance</li>
</ul>
<p>We didn't want any of those. A VPN on a kid's phone is fragile and invasive, MDM is built for companies, not families, and reading every message breaks trust with the teen.</p>
<h2>The approach: keep AI in Safari, where an extension can help</h2>
<p>Apple gives parents a sanctioned toolset through Screen Time: the <strong>FamilyControls</strong>, <strong>ManagedSettings</strong> and <strong>DeviceActivity</strong> frameworks. With the parent's approval through Family Sharing, an app can:</p>
<ol>
<li>Ask for Screen Time authorization for a child's device</li>
<li>Let the parent pick which apps to restrict, using Apple's own picker (the app only gets opaque tokens, never a list of what's installed)</li>
<li>Apply those restrictions with ManagedSettings</li>
</ol>
<p>So instead of trying to see inside AI apps, we flip the problem around. The parent restricts the standalone AI chatbot apps on the teen's iPhone, and AI stays available in Safari. In Safari, a <strong>Safari web extension</strong> can run on the chatbot sites themselves, with the teen knowing it's there.</p>
<p>That gives us a few things we cared about:</p>
<ul>
<li><strong>No VPN, no configuration profile, no MDM.</strong> Everything uses Apple's own family APIs.</li>
<li><strong>The teen still has AI.</strong> Homework help and curiosity aren't blocked. AI just happens in the browser.</li>
<li><strong>The parent approves it</strong>, through Family Sharing, the same way they'd set other Screen Time limits.</li>
</ul>
<h2>What the extension does with a prompt</h2>
<p>This is the part we were most careful about. When the teen sends a prompt on a supported AI site, the extension sends it to be classified into a <strong>category, topic and concern level</strong>, and then the prompt is discarded.</p>
<p>The parent sees things like "a chat about friendships, low concern" or a heads-up when something may need a closer look. They never see the words their kid typed. Self-harm is always watched, and parents can choose a few extra topics that matter to their family.</p>
<p>We think that's the right trade: enough signal for a parent to know when to start a conversation, without turning their teen's private questions into a transcript.</p>
<h2>Things we learned</h2>
<ul>
<li><strong>Screen Time authorization is the gate for everything.</strong> Design your onboarding around it, because a parent who skips it gets nothing.</li>
<li><strong>Tokens, not app lists.</strong> FamilyControls deliberately hides which apps are installed. Build your UI around Apple's picker rather than fighting it.</li>
<li><strong>Be honest with the kid.</strong> The teen knows TeenCue is installed. That makes the whole thing a conversation starter instead of a spy tool.</li>
<li><strong>Test with old accounts, not just fresh signups.</strong> We shipped a sign-in bug that only hit accounts in a specific state, because every test used a brand-new account. Our release checklist now includes existing accounts on every build.</li>
</ul>
<h2>What's next</h2>
<p>We'll write more about classifying prompts by topic and concern level, and about shipping one extension to Chrome, Edge, Firefox and Brave.</p>
<p>If you're building anything with Apple's Screen Time APIs, I'd love to compare notes in the comments.</p>
<p><em>TeenCue is parental controls for AI, on iPhone, iPad, Chrome, Edge, Firefox and Brave. Learn more at <a href="https://teencue.com">teencue.com</a>.</em></p>
]]></content:encoded></item></channel></rss>