Similiarity is not enough. The degree of similarity between the two works needs to be "striking or substantial".

There is no exact formula for what is considered substantial, but merely emulating the style of an artist is insufficient. It is not impossible for an AI to create a copyright violation, and future, large models may create real risks, but if you consider the very tiny model size used by stable diffusion, the presence of sufficient information about a prior work of art to create a substantial copy, would be extremely unlikely - It is a bit like asking a normal person to reproduce some work of art that they have seen years ago from memory alone - our memory is much larger than stable diffusion but still too blurry in order to provide anything that would amount to be a copyright violation. I have tried reproducing very famous paintings with stable diffusion, and in all cases the results were substantially different from the original.

Of course this discussion is highly simplified, and there can be broad protections for things such as fictional chracters, which surely can be produced by stable diffusion - so I would definitely advise consulting with a lawyer, as you suggested.