How can we stop AI remote-control clients from being installed on production servers?

November 17, 2026 5 answers
Asked by v.637!brindlevisor

Developers keep installing AI remote-control clients on production servers, and the CPU and memory use is affecting other users. Killing the processes and deleting the executables has not stopped them from coming back.

Most servers run Windows Server, with some Ubuntu. Developers need permissions to deploy code, so what controls would help without breaking deployments?

Answers

4a!quorvani_683

Start with who is allowed to change production and who owns the resulting risk. If people can restore access or reinstall software using their existing group permissions, this is not just a process cleanup problem.

Bring the resource impact and privacy concerns to the service owner and management. Agree on an approved software process and a way to enforce it, including when senior developers are involved. Otherwise you may be stuck fighting the same installation indefinitely.

Reply to this answer
Rovdali-8m

Separate deployment rights from general server administration. Deploying an approved build does not necessarily require standing local admin access or permission to run arbitrary tools on the host.

One option is to promote tested builds through a controlled deployment pipeline. For production troubleshooting that really does need elevated access, use time-limited privileges and audit the actions. Review the group memberships that currently let users undo restrictions, too.

v.637!brindlevisor:

That distinction helps. We still need a workable deployment path, so removing access without replacing it would be difficult.

Reply to this answer
_y38+clatterfob.6

Application control is worth a look on Windows Server. AppLocker can restrict what is allowed to run based on rules such as publisher or file location. An allowlist of approved software is generally more useful than chasing one executable name.

Test the rules in audit mode first and check the results against deployment tools and services before enforcing them. Also protect the policy and approved locations from modification by the same accounts you are trying to restrict. Otherwise the rule is easy to work around.

Reply to this answer
~c4!socketladle+

Check outbound traffic as well. A controlled proxy or gateway can restrict where production servers connect, and logs can help identify which hosts are reaching the remote service.

That is a separate layer from application control, not a complete fix. Check dependencies before blocking destinations so you do not take out an unrelated service.

Reply to this answer
Mivaro

For immediate containment, use your server management tooling to inventory the processes and services, alert on them, and remove them where you have authorization. That beats manually hunting across servers.

Treat it as a temporary measure though. If the same users can reinstall the clients, a recurring cleanup script just turns this into a contest.

Reply to this answer

Add your answer

Answers are reviewed before publication.

Related Discussions