What does the recent guidance on AI use in practice say about which AI models are considered safe, and does it address Section 7216?

The guidance did not spell out a specific list of approved models, but it did state that practitioners should use 'secure enterprise approved AI.' This suggests firms should be using enterprise-level AI products rather than the free, public versions of tools like ChatGPT or Claude. Practically, this means firms need to be set up properly: having a Data Processing Agreement (DPA) on file, and where possible obtaining zero data retention from the provider (each provider has its own process for this). The guidance did not directly address Section 7216 — it remains a gray area whether using these models constitutes an 'auxiliary service' versus a more substantive service when doing tax work. The consensus was that this is a positive sign regulators are finally acknowledging AI use, and that more detailed guidance is likely coming. In the meantime, members were encouraged to contact their state CPA societies (some, like California and New York, have already addressed AI) and share responses so the group can build a shared knowledge base of how different states are approaching this.