Second, the interactions here appear to be entirely steric (that means size/shape related). One of the advantages of natural antibodies is that they can bind their targets via H-bonding, ionic interactions, van der Waals interactions, and hydrophobic interactions in addition to steric interactions.
Third, clearance by the liver runs the risk of just having all of the toxin concentrated in the liver, and it's not clear that the toxins will remain bound to the bead until they are inactivated. Certainly, if we're talking about a snake-venom which is a neurotoxin, and the "down-side" to taking this and not dying is that you need a liver transplant, I think it would still be useful, but this is not any sort of panacea I would think.
Edit: Also, it seems like the "imprinting" would require a fairly large interaction surface with the target molecule. This means that its probably not even very useful for molecular biology work, since natural antibodies can distinguish between proteins with very small differences and only require half-a-dozen or so amino acids for a match. Really, I think the lesson of this article should be just how amazingly cool natural antibodies are, and that it's taken until now to maybe come up with something that could possibly replace one of these uses.