I'm wondering about the considerations when it comes to sending customer personally identifiable information (PII) to large language models (LLMs) hosted on Bedrock. Should enterprises feel secure doing this? If the info is being utilized for its intended purpose, would it typically align with privacy policies in most countries? I'm also looking for any insights from those with enterprise experience in regulated industries or any related AWS documentation on this topic.
3 Answers
You can definitely implement guardrails in Bedrock, but when it comes to privacy policies, especially across different regions, it's best to reach out to your legal team for guidance.
To start off, I'd recommend checking out the AWS documentation on data protection for Bedrock; they clearly state that they don't store or log your prompts and completions. This is crucial because it means your data won't be used for training their models or shared with third parties. If you can safely send your data to Bedrock, it shouldn't be an issue.
You really need to provide more context for a thorough answer. What jurisdiction are you dealing with? What kind of PII is involved? What specific regulated space are you in? These details matter a lot!
Related Questions
Biggest Problem With Suno AI Audio
How to Build a Custom GPT Journalist That Posts Directly to WordPress