An AI engineer has publicly urged Sam Altman to prioritize fundamental data purification over reactive guardrails for AI safety, arguing that contaminated training data inevitably breeds harmful outputs. The letter compares guardrails to patching a leaky dam, asserting that advanced AI will circumvent them, making foundational data cleansing essential.