AI Safety and Ethics: What Every Young Person Should Know

I'm Parikshet. I'm 11. I use AI every day. And I think the most important skill in my generation is not knowing how to use AI — it is knowing how to think critically about it. Here are the things I wish every kid my age understood about AI safety and ethics.

AI Bias: The Problem Is in the Data

Every AI model learns from data. If that data reflects historical human biases, the AI learns those biases too — and then enforces them at scale.

A real example: Amazon built a hiring AI trained on ten years of CVs from successful hires. Because their tech workforce had historically been mostly male, the AI learned to downgrade CVs that contained words like "women's" (as in "women's chess club"). It was systematically discriminating against women's applications and nobody noticed for two years. Amazon scrapped it when they discovered the problem in 2018.

A healthcare example: an algorithm used across US hospitals to identify patients who needed extra care was found to systematically underestimate how sick Black patients were compared to white patients — because it used healthcare spending as a proxy for health need, and Black patients had historically spent less on healthcare due to access disparities, not better health.

Bias in training data becomes bias in real-world decisions that affect real people's lives.

Deepfakes: When You Cannot Trust What You See

Deepfake technology can generate a convincing video of any person saying anything. It requires only a few minutes of real footage and a consumer-grade computer. In 2023, multiple governments and schools reported deepfake audio recordings of teachers and principals saying inappropriate things — created by students to harass staff.

Political deepfakes are more dangerous. A convincing deepfake of a world leader announcing a crisis could cause real panic before fact-checkers could respond. Detection tools exist but are in an arms race with generation tools.

My rule: if a video shows something shocking and it was just shared widely and not verified by multiple news outlets — assume it might be fake until proven otherwise.

AI in Decisions That Affect Your Life

AI systems are already making or influencing decisions about:

• Who gets a bank loan
• Who gets bail and who stays in jail pre-trial (COMPAS algorithm in the US)
• Which university applications get flagged as high-risk
• What content you see on social media (affecting your worldview)
• Whether your CV gets to a human recruiter or not

Many of these systems are proprietary — the companies or institutions using them do not have to explain their decisions. This lack of transparency is one of the central AI ethics debates happening right now. The EU's AI Act (2024) is the first major law requiring transparency for "high-risk" AI uses.

Surveillance AI

Facial recognition AI can identify individuals from CCTV footage. This has genuine safety uses — finding missing persons, identifying criminals at large. It also has deeply troubling applications — mass surveillance of protests, tracking ethnic minorities, identifying political activists in authoritarian states.

In China, the Social Credit System uses AI and surveillance cameras to track citizen behaviour and assign scores. Multiple Western cities have banned or restricted police use of facial recognition after studies found error rates were significantly higher for dark-skinned faces — meaning innocent people were more likely to be misidentified.

What AI Safety Researchers Actually Do

AI safety research asks: as AI systems become more capable, how do we ensure they remain under human control and do not cause unintended harm?

Specific problems they work on: alignment (making AI want what humans want), interpretability (understanding why AI makes its decisions), robustness (making sure AI does not fail badly in situations it hasn't seen before), and containment (how to limit what a highly capable AI can do if something goes wrong).

Anthropic (who built Claude, the AI I often use), DeepMind, and several university labs have dedicated AI safety teams. This is a field you could work in — and it matters enormously for the world my generation will live in.

What You Can Do Right Now

1. Question AI outputs. Never copy-paste without fact-checking.
2. Learn to spot deepfakes. Look for: unnatural eye blinking, hair that blurs at edges, lighting inconsistencies, audio that does not quite match lip movement.
3. Understand your algorithmic feeds. The social media algorithm is not showing you the most important content — it is showing you the content you are most likely to engage with. Those are often different things.
4. If AI makes a decision about you that feels wrong, ask for human review. In many countries you have legal rights to do this.
5. Talk about this. The kids who understand AI ethics will shape the policies that govern AI. Start now.