I agree (mostly) with Andrew's answer because it is a common misconception about how accountabilites work. However, part of the reason this topic is confusing, even for experienced Holacracy practitioners, is because there are two important caveats to this.
1) If someone is consciously taking action on an ongoing basis outside of a role they fill, but a defined role already exists, then they can only do that if they believe their action would resolve or prevent more tension for the organization than it would likely create (4.3.1.b). This is covered under the rules for Individual Action.
Additionally, they have a duty to explain their action and the intent behind it to any partner who fills a role that may be significantly impacted. So, yes, if the Customer Service Systems (CSS) role felt a tension about someone else doing that work, they could propose governance to restrict that, but they could also just tell them to stop. This is covered in 4.3.2. So, unless the person is doing the work of another role unknowingly (or they don't think there is any impact), the burden is actually on the person taking Individual Action to explain themselves.
2) A smaller caveat to the "accountabilities aren't authorities" issue, is that accountabilities do give you "authority" to process specific tensions. Of course, this isn't limited to accountabilities, it also applies to purpose. In this sense, it's more like representative power. A role-filler has certain pathways available to them (objections, proposals, etc.) that aren't available to someone outside of the role.
I hope this helps!