He refers to an image, a screen capture from 7/23/08:
"Also interesting to see how the interpretations of the API levels has shifted a bit:
1 = API 0-50 = excellent (old) => good (new)
2 = API 51-100 = good => moderate
3A = API 101-150 = slightly polluted => unhealthy for sensitive groups
3B = API 151-200 = light polluted => unhealthy
4A = API 201-250 = moderate polluted => very unhealthy
4B = API 251-300 = moderate-heavy polluted => hazardous
Especially the re-classification of ‘light polluted’ to ‘unhealthy’ is remarkable; the new classification is in fact very similar to the US-EPA."
I was quite surprised to see those translations, not only what Tom pointed out, but also the re-categorization of 51-100 as "moderate" instead of "good," and the first-time use of the term "hazardous." I was additionally very surprised because, at that time, I remember that the Chinese on the same site was still the same as in the official explanation:
1 = API 0-50 = 优 "excellent"
2 = API 51-100 = 良 "good"
3A = API 101-150 = 轻微污染 "slightly polluted"
3B = API 151-200 = 轻度污染 "light polluted"
4A = API 201-250 = 中度污染 "moderate polluted"
4B = API 251-300 = 中度重污染 "moderate-heavy polluted"
5 = API 301-500 = 重污染 "heavily polluted"
In other words, for a short period of time on the Beijing Olympic air quality site, an API of 51-100 was listed as "moderate" air quality in English, but "良" (literally "good") air quality in Chinese.
Shortly thereafter, I noticed that the English changed again, this time removing the word "hazardous" for the worst level but keeping "moderate" for 51-100. Unfortunately, I don't have a screenshot of this set of translations.
As of now (8/1 evening), there is a third presentation of the air quality levels in English, and, as far as I know, for the first time we have a new presentation in Chinese as well. Avoiding terms altogether, the air quality level is simply indicated as no-context roman numerals (these are used on the Chinese version as well as the English version):
Clearly, China is preparing for hundreds (thousands?) of scientists and journalists scrutinizing its air quality numbers and judgments, and hasn't 100% settled on exactly how to present them.