Nobody (absolutely NOBODY) can say after which dose you will have a x% higher risk of getting cancer. In theory one single molecule is enough to cause the harm. Of course, that's just theory. The most practical and realistic approach to cancerogenic substances is to avoid contact at all, or as much as possible.
Yo Murphy!
Since i see quite some interest in this thread and people are already being concerned for smoking the..."godsend" spice, i think it would be good to elaborate more on this issue. For me , my prime interest is people to get an as most accurate as possible idea of how the "mechanics" of mutagenicity/carcinogenicity take place.
First of all carcinogenesis is not usually a "one step" procedure. That means that usually before a cell turns carcinic more than one changes happen to it.They have to do with cell cycle regulation (especially the cell division -mitotic procedures-) , with the cell's mechanisms to rectify this -a minimum of them must be turned "off" lets say-, and then later other steps that could possibly allow the cell to become dislogged from the tissue it resides in, migrate to another position and be CAPABLE of angiogenesis to ensure a supply of blood to the tumor ,and hence its survival. The last series of events concern "metastasis" ,a phenomenon when cancer cells migrate to multiple organs usually through the bloodflow.This stage also is where the quite promising "antimetastatic" medications fit in ,inhibiting angiogenesis (blood vessel formation) . All of the above mechanisms are controled at a genetic level, hence mutations might cause cancer IF they hit the "right" genes. Now this "IF" is the whole story...
Some people have predisposition to cancer meaning that propably they carry mutations/alleles and to put it simply they are "one step ahead" in cancer formation than those who dont.
On the "IF" issue : Indeed a single molecule of a substance COULD cause a mutation as its DIRECT action or as action of one of its metabolites. What would have to happen is the molecule or the metabolite to interact with DNA and -through a variety of mechanisms- either interfere with its correct duplication during a mitotic cycle or straightforwardly change something in its chemical structure.
I would say this issue looks akin to the radioactivity "bullet and target" mutagenicity : Imagine the mutagenic molecule as a "stray bullet" in a room.It might hit noone.It might hit someone in a multitude of ways from crippling him to only a superficial wound to killing him by a "bullseye" on his/her heart. So a stray bullet can be dangerous. Increasing the QUANTITY of the molecules ,it is as if increasing the quantity of stray bullets. Increasing the FREQUENCY of the administration,increases via propability that in a "bullet firing session" someone gets hurt.
So as said before , the safest option would be not to fire straight bullets in a room ,there is chance someone dropping dead or worse

. If one HAS to fire bullets -bulletophiliacs

- then fewer bullets, less frequent firings guarantee
comparative safety than a barrage of bullets.
Hope this clears up the issue a bit, and mostly to clear up WHY cant anyone tell what is the chance of causing mutation to DNA AND lead to carcinogenicity with certainity.
Can anyone comment on reliability and reproducibility of the mentioned "GreenScreen HC"-assay?
I cannot promise anything yet, but if i can have a toxicologist look at the data i will do and report back. I will have a look at them myself, although im far from qualified on this field