Pillar 02

Risks, in proportion

We take risk seriously without taking it apocalyptically. Each entry names the harm, the evidence, and the mitigations.

Misinformation at scale

Cheaper synthetic media changes the cost curve. Mitigations: provenance, platform policy, media literacy.

Labor displacement

Task automation hits unevenly. Real questions are about transition, not extinction. Track displacement and creation together.

Bias & fairness

Models inherit data. Mitigations: evaluation suites, dataset transparency, redress mechanisms.

Concentration of power

Compute and capital concentrate. Open weights, antitrust, and public infrastructure matter.

Security & misuse

Phishing, fraud, cyber. Real but tractable — defenders also benefit from AI.

Environmental cost

Training and inference draw energy and water. Track efficiency gains alongside absolute use.

Our editorial line. We avoid two failure modes: dismissing real harms ("everything is fine") and apocalyptic framing ("everything is over"). Both make readers less able to act.