[Paul_Diraq]

The article mentions activities of 1 million bequerel per liter. That means every liter produces 1E6 helium atoms per second. There are about 1E9 liter of water (one million tons) and a year has about 3E7 seconds. So one could produce about 3E22 helium atoms per year or 0.05 mol (one mol is 6E23 atoms) or 0.2 grams or about one liter (density 1.79 g/L).

This is a back of the envelop calculation, but realistically you would have to deal with leakage so you get much less than a liter of helium.

[ggm]

Name checks out. Assuming you didn't drop a power or two, it illustrates the huge gap between useful quantities for their physical or chemical properties, and the astoundingly small amounts of radioactive nucleotides which cause contamination.