AI Safety in Your Life
AI safety can sound like a distant, abstract concern — something that matters for billion-dollar research labs and government policy meetings, but not for a middle school student. That impression is wrong. AI systems are already woven into the texture of daily life for most young people today, and the ways those systems can go wrong affect you directly. Seeing the connection clearly makes AI safety feel less abstract and more urgent.
Recommendation Algorithms: The Invisible Curator
Every time you use a video platform, social media app, music service, or news aggregator, an AI recommendation algorithm decides what you see next. These algorithms are optimized to maximize engagement — to keep you watching, scrolling, and interacting as long as possible. The safety concern here is subtle but real. A misaligned recommendation algorithm might learn that outrage, anxiety, or sensational content keeps you engaged longer than calm, accurate, balanced content. If so, it will show you more of the emotionally intense content — not because anyone programmed it to, but because that is what the optimization discovered. Over months and years, this can distort what a person believes is normal, true, or important. You do not have to be a passive subject of recommendation algorithms. Actively searching for content rather than following recommendations, consciously seeking out diverse viewpoints, and setting intentional time limits are all forms of AI safety in practice.
Engagement optimization means an algorithm is tuned to maximize how long users spend on a platform. This can produce misalignment: the AI does its job perfectly from the metric standpoint while making users more anxious, more polarized, or less accurately informed.
AI Homework Helpers: Helpful Tool, Real Risks
AI writing and tutoring tools can be genuinely useful for learning. They can explain concepts, give feedback on drafts, and help you get unstuck. But they carry specific risks for students. Hallucination is the biggest one. If you ask an AI assistant to explain a historical event and it confidently describes details that are wrong, and you write those details into your paper, you have spread misinformation without knowing it. Verification is essential. There is also the risk of using AI to do work that is supposed to develop your own skills. Submitting AI-generated text as your own may feel like a shortcut, but it bypasses the actual learning process — meaning you miss the skill development, not just the grade. This is a personal safety concern as much as an academic integrity one: the skills you build now matter for your future.
Voice Assistants and Smart Devices
Smart speakers, phone assistants, and other AI-powered devices that listen for voice commands raise genuine privacy questions. When does the device start recording? What happens to those recordings? Who can access them? Most responsible manufacturers publish privacy policies that describe these practices, but policies can change and are often long and complex. Good habits include knowing how to turn off or mute listening devices when you want private conversations, understanding the settings on devices you use, and being aware that smart devices in your home may be collecting data about your household's routines.
Deepfakes and Synthetic Media
AI can generate realistic fake images, videos, and audio of real people saying or doing things they never said or did. This technology is called deepfakes. At the social level, deepfakes threaten to make it harder to trust video evidence. At the personal level, they can be used to harass individuals by creating fake images of them without consent. Recognizing deepfakes is increasingly difficult as the technology improves. Habits that help: be skeptical of surprising or sensational video and audio content, look for independent news coverage before sharing, and report content that appears to be non-consensual synthetic media of real people.
Creating synthetic images, videos, or audio of a real person without their consent — especially content intended to embarrass or harass them — is a serious ethical violation and in many places is illegal. This applies to peers, public figures, and everyone in between.
Match each AI system in daily life to its primary safety concern.
Terms
Definitions
Drag terms onto their definitions, or click a term then click a definition to match.
A recommendation algorithm keeps showing a user increasingly extreme content about a topic because extreme content generates more comments and watch time. Which AI failure type does this best illustrate?
Why is it risky to include AI-generated statistics in a school report without checking them first?
One Day, Five AI Touchpoints
- Step 1: For the next twenty-four hours (or reconstruct yesterday), keep a log of every time you interact with an AI system — recommendation feeds, autocomplete, voice assistants, maps, search, anything.
- Step 2: Record at least five AI touchpoints. For each one, note: What task did the AI perform? Who built this system? What could go wrong if this AI made a mistake?
- Step 3: For two of your touchpoints, identify which of the five failure types from Lesson 3 (bias, brittleness, hallucination, misalignment, privacy risk) could most plausibly apply.
- Step 4: Identify one touchpoint where you already use a safe habit (verification, privacy protection, involving a human) without thinking about it.
- Step 5: Identify one touchpoint where you could start using a safer habit. Write a specific action you will take.