What changed on the page
New section titled “Our AI governance principles” states that Plaid builds models “through structured, repeatable workflows,” uses “legal review and benchmark testing,” and monitors models for “data quality, latency, reliability, and signs of drift or degradation.”