Get the synthesize_data() function I had Claude write: synthesize-data-function.R on GitHub.
The Posit blog post I show in the video can be found here.
Two other good general links about data security when working LLMs are from Mintivite and datastudios.org.
Transcript
Click on the transcript to go to that point in the video. Please note that transcripts are auto generated and may contain minor inaccuracies.
Loading transcript...
I walked through what it means to give an LLM access to your data, since whether that data stays private doesn't have a single answer. I covered provider settings, zero data retention agreements, synthetic data, and local models, and I mention a synthetic data function I had Claude write during an earlier livestream.
You'll learn:
- Why keeping your data hidden from an AI agent limits how useful it can be, and what that trade-off looks like
- How to find and turn off the setting that stops Anthropic from using your Claude data to improve its models
- What a zero data retention agreement is and two ways to get one: an enterprise account or Posit Assistant
- How to use a synthetic copy of your data by generating it in a separate project with plain R code before bringing it into the project Claude Code works in
- The trade-offs of local models, including weaker performance than frontier models and the need for a powerful computer
Tools mentioned: ChatGPT, Claude, Claude Code, Claude Opus 5.5, Posit Assistant, OpenAI models
Key moments:
- 00:03 Can your data stay private when using LLMs
- 00:47 Downside of hiding your data from an AI agent
- 01:18 Top models run on cloud servers
- 01:53 Claude's 30-day default and the model improvement setting
- 02:36 Zero data retention agreements
- 03:11 Posit Assistant and zero data retention
- 03:37 Making a synthetic copy of your data
- 04:46 Local models and their limits
