Sen. Mark Warner Raises Concerns About Google’s Medical AI Tool

Sen. Mark Warner has questioned the readiness of Google’s medical artificial intelligence tool, citing reports of repeated inaccuracies and concerns by company researchers.

In a letter to Google on Tuesday, he said the premature rollout of the Med-PaLM 2 large language model could diminish trust in medical institutions, worsen racial differences in health outcomes and increase the risk of diagnostic and care errors.

Med-PaLM 2 answers consumer health questions on behalf of providers. According to Google, the AI is 86.5 percent accurate when responding to U.S. Medical License Exam-style queries.

A limited number of hospitals, including Baton Rouge General in Louisiana and the Mayo Clinic, are testing Med-PaLM 2.

Warner enclosed a list of questions in the letter, including one that would require Google executives to explain the AI tool’s data collection and retention processes.

Google maintained that Med-PaLM 2 allows health care organizations to retain control over data. The company said it is committed to ensuring privacy and equity as it develops the technology, The Hill reported Tuesday.