Yeah I could see it being a problem if you're doing work on closed source or repos with sensitive credentials. Since my usage has all been on open source projects I'd be happy to share everything I'm doing if it can help train better models.