What should I consider when sending customer personally identifiable information (PII) to LLMs hosted on Bedrock? Are enterprises generally comfortable with this practice, and does using such information align with privacy policies across different countries? I'm looking for insights from anyone with enterprise experience in a regulated sector, as well as any relevant AWS documentation on this topic.
3 Answers
You can set up safeguards within Bedrock, but keep in mind you’ll want to check with your legal team regarding privacy policies in your specific region. Different areas have different regulations, so it's best to ensure compliance from the get-go.
To really dive deep into this, I'd need more specifics about your situation. What jurisdiction are you in? What type of PII are we discussing? Each regulated space can have different requirements.
It's a good idea to start with AWS documentation on this. They mention that Amazon Bedrock doesn't store or log your prompts and completions, which is reassuring. They also state that they don’t use this information to train their models or share it with third parties. As long as the data is securely and privately transmitted to Bedrock, I think you should be fine.
Related Questions
Neural Network Simulation Tool
xAI Grok Token Calculator
DeepSeek Token Calculator
Google Gemini Token Calculator
Meta LLaMA Token Calculator
OpenAI Token Calculator