Is it safe to install Claude Desktop on a machine that has access to client files, and how do you get IT/compliance on board?

One member said their IT group has not gotten comfortable with installing Claude Desktop on machines connected to client files, so they've been limited to a sandboxed laptop with no access to real data for experimentation, and can't yet deploy on real client data. They noted they need to convince IT that the risk can be managed with proper settings. Another member responded that most of these risks can likely be mitigated, and suggested it could be worth polling others (e.g., on social media) about their policies, pointing out that many everyday tools already have AI features embedded, raising the question of why Claude specifically draws more scrutiny. This point led into a discussion of IRC 7216 concerns without a full resolution.