How should we handle requests to run AI-generated Python scripts?

November 18, 2026 5 answers
Asked by Rell

Nontechnical staff are asking for software to run scripts generated with AI. We want to support useful automation, but the code, dependencies, and access involved need some scrutiny. How are people approaching this without just banning everything?

Answers

_a47!veldriko.5

Start with the task, not the software request. Ask what the script needs to do, what data it will touch, and who owns the result if it goes wrong. A read-only report over limited data is a different proposition from a script that changes shared files or business records.

Have someone who understands Python review the script and its dependencies before it runs. Keep the approval tied to a specific version, so a changed script goes back for review. A short, accessible request process gives staff a route to get help instead of leaving them to work around IT.

Rell:

Should the business owner be involved as well as the code reviewer?

m.642!zindaro:

Yes. The reviewer can assess the code, but the business owner should be able to explain the need, approve the data access, and take responsibility for the outcome.

Reply to this answer
5a~wobblecaddy_746

Give experimentation its own environment rather than installing tools everywhere by default. Limit the account's permissions, data access, and network access to what the task needs, and keep it away from production systems where possible. That reduces the damage a bad script can do, though it doesn't make unreviewed code safe.

Also look at the packages. AI-generated instructions can suggest dependencies without checking whether they're suitable. Review them and use an approved source rather than letting a script install whatever it asks for.

_d42!norlixa.8:

A virtual environment helps separate Python dependencies, but it isn't a security sandbox. You still need controls around what the process can access.

Reply to this answer
c.738!quorlavi

I would say no to running arbitrary generated code until there's a review path. The person asking doesn't necessarily need to be a developer, but someone responsible for the script needs to understand what it does and how to maintain it. Otherwise every small change becomes another mystery.

Reply to this answer
v.391+quendavi

Separate Python from the editor request. Visual Studio isn't required just to run a Python script, and approving an editor is not the same thing as approving the code someone runs with it. Check the actual task, the tools needed, and the licensing and support costs before deciding what to install.

Reply to this answer
Nima

Don't focus only on Python. People may try other scripting tools if they can't use it, so apply the same thinking to what those tools can access and execute. Restrictions can help, but pair them with a practical way to request automation and get it reviewed.

Reply to this answer

Add your answer

Answers are reviewed before publication.

Related Discussions