Most test runners have an option (or can be easily modified) to fail when 0 or less than X tests have been run. You should use it for situations like this.
What if there's a bug where your CI ignores all steps? There's always some situation possible and we can go all the way to nasal demons. You have to accept the risk at some point.