Zero Data Retention by llms · AI Marketing Insiders

Zero Data Retention by llms

I thought I'd share this because it might be helpful to some of you, especially if you care about privacy.

So as you probably know one of the drawbacks of using LLMs, web based and API, is that almost all major ones, like OpenAi/Anthropic/Gemini etc. retain your input and output data for 30 days.

In business and even personal uses it's a problem because you don't have data privacy plus there's additional risk of your data being stolen from their servers.

So you have an option... to run a local llm, like Google's Gemma/Nvidia Nemotron/OpenAI GPT OSS/Qwen/Deepseek etc. and everything stay on your computer.

I actually tried it few days ago :) but quickly found out that you need a powerful machine to run it efficiently. Think a $15-20k workstation not my laptop lol.

It worked but a query that I do on ChatGPT for example, which took 1 minute to complete, took 65 minutes on my laptop 😄. In addition local llms are not nearly as good as API models. But they are good enough for many simpler tasks where top notch advanced reasoning is not needed.

And if you/your client can invest in a $15/20k machine to run it it's an option.

So I dug deeper and found a more affordable option :). Perplexity offers Sonar/Sonar Pro API. It has Zero Data Retention Policy. None of your inputs/outputs are retained, only metadata.

https://docs.perplexity.ai/docs/resources/privacy-security

Hope this helps!

11 comments

Zero Data Retention by llms