Large language models are being used in financial reporting, but they often produce "hallucinations" - factually incorrect information that sounds plausible. This can lead to significant risks, including SEC enforcement and reputational harm, and requires careful deployment and oversight to mitigate.