AI Ethics & Society

AI ethics is no longer an academic sidebar. It’s a set of design choices you make when you build, deploy, or buy AI. Four questions to hold in mind.

1. Labor — whose work is being displaced?

The World Economic Forum projects 40% of global job skills will shift by 2030 due to AI. Generative AI has disproportionate impact on knowledge-worker entry-level roles: paralegals, junior copywriters, customer support, junior developers. If you’re building AI products, this isn’t theoretical — your product is somebody’s job.

2. Copyright — what’s the training data story?

Active litigation is redefining the boundary. The New York Times v. OpenAI, artists v. Stability, Getty v. Stability, and ongoing EU AI Act implementation will set the 2026–28 compliance floor. For your own training data (if you fine-tune): document provenance, respect opt-outs, and get comfortable with “we don’t know yet” as a legal answer.

3. Misinformation — generative content at scale

A high-school senior with Runway and ElevenLabs can produce a convincing fake video in an hour. Political campaigns, scammers, and state actors have all used generative tools in the wild. Media literacy isn’t a soft skill anymore; it’s civic infrastructure. If you build generation tools, build provenance watermarks in.

4. Concentration of power

Three US companies — OpenAI, Anthropic, Google — plus a handful of Chinese labs, control the frontier. The capital required to train frontier models is tens of billions of dollars. Open models from Meta, Mistral, and Qwen are the counter-pressure, but the oligopoly question is real. It shapes what kinds of AI we’ll get and who gets to build them.

Your obligations as a builder

• Label AI-generated content.
• Red-team your product with people who are likely to be harmed.
• Measure error rates across demographic groups, not just overall.
• Design for dignity: assume your product will be used against someone, and make that harder.