Your Data is Unique
Many Solutions claim they can achieve near-total redaction.
In reality, most of these systems were only tested against publicly available datasets, such as I2B2 or MIMIC, and fail to deliver the same on your data. Everyone’s data has different edge cases, making the 99% threshold challenging for traditional AI.