Model-agnostic PII detection with LLMs
AWS ships model-agnostic PII detector on Bedrock—swap models, swap entity types, no retraining needed.
Why it matters
A practical pattern for practitioners: define PII detection rules in prompts rather than code, enabling rapid adaptation to new entity types and model switching without ML ops overhead. Outperforms off-the-shelf tools and nine LLM baselines across five benchmarks.
The key facts
11 to knowModel-agnostic detector works with any LLM on Amazon Bedrock
Entity types defined in prompt, not code—enables rapid reconfiguration
No retraining required to detect new entity types
Outperforms off-the-shelf PII detection tools
Outperforms nine LLM-based detectors
Evaluated across five public corpora
Model-agnostic approach: works with any LLM on Bedrock
Prompt-configurable entity detection (no retraining required)
Outperforms off-the-shelf tool and nine LLM-based detectors
Evaluated across five public PII corpora
Configurable for new entity types without model updates
Go to the source
AWS Machine Learning Blogaws.amazon.com
Publisher excerpt: A configurable, model-agnostic detector that turns any large language model on Amazon Bedrock into a PII detector. Because the entities to detect live in a prompt rather than in code, one detector adapts to new entity types without retraining, and it outperforms an off-the-shelf tool across five…