HHS Scales Back AI Translation Safeguards as California Moves to Ban Unreviewed AI in Care

*TL;DR The HHS Office for Civil Rights stopped requiring human review of AI‑generated medical translations, but California’s pending AB 1242 would bar AI‑only translations from being considered qualified, creating a split between federal flexibility and state strictness.

Context Section 1557 of the Affordable Care Act bans discrimination based on national origin, obligating health providers to offer qualified interpreters for patients with limited English proficiency. The rule defines a qualified interpreter as a person fluent in both languages and bound by professional ethics, including confidentiality. While the regulation does not expressly forbid machine translation, it mandates human review when accuracy is essential— a common scenario in clinical documents.

Key Facts - In May 2024 the HHS Office for Civil Rights (OCR) revised the 2020 § 1557 regulation, removing the requirement that every AI‑generated translation be checked by a qualified human or that patients be notified of AI use. The change came despite public comments urging stricter safeguards. - A December 2024 OCR Dear‑Colleague Letter introduced a narrow emergency exception: providers may delay human review of AI translations during urgent care, but must conduct the review “as soon as practicable” and must warn patients that the translation could contain errors. - California’s pending AB 1242 would redefine “qualified interpreter,” “qualified translator,” and “translation service” to exclude any AI output that lacks human review, effectively banning unreviewed AI from being counted as a qualified language‑access tool. - Other state measures, such as Texas’s Responsible AI Governance Act and California’s AB 3030, focus on disclosure rather than outright exclusion, highlighting a patchwork of approaches.

What It Means For health systems, the federal rollback reduces immediate compliance burdens but leaves ambiguity: many clinical texts still meet the regulation’s “critical to rights, benefits, or meaningful access” test, which could trigger the human‑review requirement. The December letter clarifies that emergencies do not grant a blanket exemption; providers must still document the delay and inform patients.

In California, AB 1242 would force hospitals and clinics to retain a human translator for any patient‑facing document, regardless of speed or cost. Organizations operating in multiple states will need dual compliance strategies—using AI as a drafting aid under federal guidance while maintaining a human‑review pipeline to satisfy California law.

Practical steps for providers include: 1. Auditing current AI translation tools for accuracy in medical terminology. 2. Establishing rapid‑turnaround human review workflows for emergency use. 3. Updating patient consent forms to disclose AI involvement where required. 4. Monitoring state legislation to adjust policies before AB 1242 takes effect.

Looking Ahead Watch for the California Senate vote on AB 1242 and for any further OCR guidance that could tighten or clarify the federal stance on AI‑driven language services.