I've seen plenty of scenarios when a test has also given a false sense of security. It's not actually testing anything important, but looks like it does and should be, but doesn't actually catch defects. This is also a case of no tests being better as at least everyone knows something's untested :/