This is a great question. Luckily the answer is simple. You still only calculate potential alcohol based on the original Brix reading. “Negative Brixes,” or when the density of your fermented solution reads below the 0.00 °Brix mark on your hydrometer, happen because they are just that: Fermented. Alcohol is much less dense than water and when we come to the end of the fermentation and the sugar has almost all been converted to alcohol, Brix is truly only a measurement of solution density. As the relatively more-dense water and sugar solution (grape juice) gets transformed into a water and ethanol solution (wine) the new alcohol content skews the results artificially lower. There is a significant contribution from the alcohol to the ending Brix reading, which means that by the end of fermentation the Brix “reading” itself is artificially low. This means that when calculating potential alcohol we do not count negative Brix degrees as actually contributing to the alcohol level. In a way, they are like “phantom Brix units” that are really meaningless.
Estimating potential alcohol is tough, even for veteran winemakers. Things like yeast strain, fermentation temperature and dried up grapes can make it hard to first get an accurate initial Brix reading and to subsequently translate an initial Brix reading into final alcohol. Though in a lab we may say that for every degree Brix you’ll get X amount of ethanol, in reality it’s never that simple. You can certainly help yourself, though, by only using your hydrometer to help you calculate potential alcohol at the beginning of the fermentation, where it will most accurately reflect what you’ll be getting from the sugar in your juice or must.