It is not circular because the Pythagorean theorem can be verified separately (by God knows how many methods of your choice: there are books full of proofs just for the Pythagorean theorem).
It might be easier to think of the law of cosines, not as permitting a special case of the Pythagorean theorem, so much as it is a generalization of the Pythagorean theorem to triangles that aren't necessarily right.
The law might subsume the Pythagorean theorem, in other words, but that doesn't mean the Pythagorean theorem itself can't be validated in other ways.