Lock it Down - Privacy with AI Tools

🧭 THIS WEEK AT AI SECOND ACT

Howdy, now we’re into the semi-boring but useful area of security and privacy with AI! Read below for brief details on how to keep your information private as needed. In short, check the settings of your tool of choice and decide what to share or not share.

👉 Hit "Reply" and let me know what you want more (or less) of, or use the poll to give feedback. My goal is to make this as valuable and practical as possible as we navigate the new AI era. 🚀

🧰 AI NEWS + LEARNING

Here’s a few things I found recently:

• Huge Hugging Face learning resources - lazy poolside/beachside, summer reading and learning!

• Amazon launched a new ‘vibe coding’ tool, Kiro, but unique in that it goes back into the good ‘old days of SDLC and does ‘crazy’ things like write requirements! Yes, I know that’s a litlte old school.

• Google PM interview includes vibe coding! We'd better start learning!

• Mistral launched Voxtral - open-source language models.

🗺️ FEATURED INSIGHT

By default, your AI tool of choice may be ‘digesting’ your information, and using your data, your prompts, and your documents you upload to further train the AI models. The biggest AI companies want your data to continue to train their AI, as your information can add value to them!

[Remember, large language models are trained on huge amounts of data - the more nuanced and specific that data is and becomes, the more valuable the model can become with unique information, insights, and answers to user queries (prompts)]

Time to check the settings and make decisions!

The Golden Rule: The Billboard Test

This is the only rule you truly need to remember:

Never paste anything into a public AI tool that you wouldn't be comfortable seeing on a public billboard.

This includes:

• Proprietary source code

• Confidential client information

• Internal financial data

• Detailed strategic plans

• Unannounced product details

• Personal performance reviews

Of course, this general advice depends on whether you are using consumer-facing AI tools or those same or similar tools within an enterprise or business environment.

• Consumer - you use or subscribe personally to, i.e., ChatGPT - be very wary of the privacy settings and policies.

•  Business - if within a business/enterprise environment, the good news is that likely your IT department has gone to great lengths to minimise/eliminate privacy concerns. Check the guidelines internally.

Priority #1 is to stop sharing chats to the mode/tool.

1. ChatGPT —> Settings —> Data Controls —> ‘Improve the model for everyone’ —> OFF!!

2. Anthropic Claude - Privacy built in, yay. - But, be careful of conversations you give feedback on, or simply don’t do that!

3. Google - not so clear in the documentation I read, but go here for settings.

For whatever tool you’re using, check the settings and disable any sharing or learning, or ‘improving’ of the model.

Next → Understand the IT situation and Privacy Issues

• if a consumer-facing tool, double-check all the settings to make sure you understand what is and is not being shared.

• Business/enterprise situation - should be more forgiving based on the policies of the IT department.

Consider Shared Information

• Instead of super-specific names, financial figures, get into the habit of sanitization.

• The Sanitization Habit: Get the Help, Keep the Secret

  So what if you need help with a sensitive document? You don't paste the real thing. You sanitize it first.

  Create a generic version by replacing all sensitive information with placeholders.

  Example: Drafting a performance improvement plan.

• Don't Paste This:

• Paste This Instead:

"Draft a performance improvement plan for a senior sales employee. Their recent quarterly performance was approximately 35% of their target. The key areas for improvement are in generating new client leads and converting them to sales. The employee is experienced but has shown a decline in these specific metrics."

In short, instead of super-specific, using names and actual data/figures, AI chatbot tools can still help you with more sanitized prompts.

Delete chats, memories & shared information

The settings of these tools allow you to delete chats, either individually or all. Also you can delete and created ‘memories’, and files uploaded. This can be a good way to restart and simply reset.

Check the settings of your tool(s) of choice, decide what to share, and understand that these large companies want your information!

Future-proof your career with AI
— Brett

👉 Hit “Reply” and share your experience — I read every one!

_{Picture by FlyD on Unsplash.}