What did the IRS Office of Professional Responsibility (OPR) say today about AI use by practitioners?

The OPR (a body within the IRS focused on practitioner standards, not the IRS itself) released guidance on AI centered on Circular 230. Nothing in it was described as revolutionary — it largely restates existing obligations: don't rely on AI output blindly, review and vet it, understand where it sources information, and substantiate anything before relying on it. A key emphasis was practitioner competence — being expected to understand how these models actually work (e.g., LLMs predict the next token from training data, which is a source of hallucinations). The guidance also touched on firm responsibilities if you have staff: proper procedures, training, an updated AI policy, and updated WSP (written supervisory procedures) reflecting how AI is and isn't used. Written advice must still be grounded in accurate, real authority. Finally, the guidance addressed fees, suggesting that if AI reduces the time spent on client work, fees charged (especially hourly) should reflect that reduction — a position one speaker pushed back on, arguing reviewing AI output can be just as time-consuming as doing the work directly, so time savings aren't necessarily the right standard.