BMI is nothing more than mass divided by height squared — a ratio the Belgian mathematician Adolphe Quetelet devised in the 1830s while studying the "average man" across a population, not any individual patient. The World Health Organization adopted it as a population-level screening tool because it's cheap, needs no equipment beyond scales and a tape measure, and correlates reasonably well with body fat percentage across a population — its well-known weakness is that it says nothing about any one person's actual composition of muscle versus fat.
A person at 70 kg and 175 cm has a BMI of 22.9 — squarely in the "Normal weight" band (18.5–24.9). At the same height, crossing into "Overweight" (BMI ≥ 25) happens at 76.6 kg, and "Obese" (BMI ≥ 30) at 91.9 kg — the chart above traces those exact boundaries across every height, not just this one.
Body mass scales roughly with volume, and volume scales with the cube of a linear dimension like height — but weight-bearing area and practical measurement made height-squared the better-fitting compromise Adolphe Quetelet settled on in the 1830s. It's an empirical fit to population data, not a law of physics.
BMI only ever sees total mass and height — it has no way to distinguish a kilogram of muscle from a kilogram of fat. A muscular, lean athlete can post an "overweight" BMI despite having very low body fat, which is the single most common false positive the number produces.
Reasonably well within the normal adult height range, which is why it remains useful — but it systematically overestimates fatness in very tall people and underestimates it in very short people, a known distortion from the height-squared approximation not being perfect.
No — the WHO standard thresholds shown here are calibrated for adults of European descent. Several national health bodies use lower overweight/obese cutoffs for South Asian, Southeast Asian and Chinese populations, since health risk at a given BMI is measurably higher in those groups at the same number.