How to Keep Your Finance Data Secure when Using AI [5 Step Framework]

Apr 05, 2024

Read time - 4 mins

If you're worried your data's at risk when using AI, keep reading.



I'm generally quite paranoid.

A family I once knew where I grew up recently had their entire identities stolen.

My kids talk to Alexa (who I swear listens when she shouldn't).

And, I'm sure I see Instagram ads that relate to private voice conversations.

Plus, that's just personal data!

So I'm super vigilant when it comes to sharing company data.

And you and your team should be too…



Data Security


Data security is about so much more than 'data being used to train AI models'.

Granted, it's not nice to think that company data is being used to train an algorithm.

But there are other things to look for, and other safeguards you can put in place to make doubly sure your data is safe.

Fancy a data security framework?



The Framework


1. Compliance


The first thing to look for is compliance standards.

  • Are they SOC2 compliant?
  • Do they have an ISO certifications?
  • Do they comply with GDPR, CCPA, HIPAA?

To see examples visit, https://trust.anthropic.com/ and https://trust.openai.com/

NOTE - OpenAI's trust centre only applies to their Team and Enterprise plans.

Key Takeaway - If you can't see a badge or two, you should probably look elsewhere.

2. Policy


Once you've validated that the platform is compliant with various standards. It's time to review their respective privacy policies and terms of use.

As with all of these things, the devil's in the detail.

PRO Tip - Find the relevant policy and use Ctrl + F to find key words like 'Data' or 'Training'. This should shortcut you to the most important areas of the document.

Key Takeaway - Compliance standards are only half the battle. Make sure you always review the terms.

3. Logins


Many platforms now offer the ability to sign in using SSO (Single Sign On) using a Google or a Microsoft account.

This is generally seen as safe. But if you're using one login for everything, this means you have a single point of failure.

Plus, if you're using a personal email it's likely your personal data will be shared with the platform you're logging into.

Another thing to consider is 2FA (2 factor authentication) which adds another authentication step via an SMS or push notification.

If you're not using 2FA, make sure it's enabled.

If you don't have this option, then using a password manager and having separate accounts for each platform with secure passwords could be a better route.

I use LastPass, but there're plenty of other options.

Key Takeaways - if you're using SSO, enable 2FA, if not, create separate accounts, and try and avoid using personal logins.

4. Training


Many of the free AI tools use your data to train the model by default. Even ChatGPT Plus (that you pay for) uses your data to train the model. You can opt out, but you lose your chat history.

Google Gemini even has a pop up to say 'Your conversations are reviewed by human reviewers, don't share anything you don't want reviewed or used'.

As far as I know right now, Claude by Anthropic and Copilot by Microsoft don't use your data to train their algorithms.

Some of you may know I'm also a fan of Poe, who have a security shield system which you can see here.

PRO Tip - You can use ChatGPT 3.5, Google Gemini & Claude for free using Poe, and your data isn't used to train the models. Just be careful, as I can't find anything about Poe's compliance standards.

Key Takeaway - Most free AI chats use your data to train their models, but there are workarounds.

5. Anonymisation


Lastly, if you're wanting to be doubly doubly doubly safe, anonymise and data you use:

  • Use fake company and individual names
  • Use the Random function in Excel
  • Redact PDF documentation

Key Takeaway - Remove anything sensitive from any data that you share.



Putting it Into Practise

  1. Set up 2FA on your accounts. Microsoft instructions here, Google instructions here.
  2. Try this prompt using https://copilot.microsoft.com/
    "Review the privacy policy on this page - https://openai.com/policies/privacy-policy/ - and outline any causes for concern that relate to data security and model training"
  3. Switch to using Copilot, Claude or Poe if you don't need to use any of ChatGPTs advanced features (like data analysis and GPTs)
  4. If you love ChatGPT, then upgrade to the Teams or Enterprise plan to ensure maximum security.

You'll be working more securely in no time.

That's it for this edition.

Speak soon

Adam

P.S. - The best way to help me is by sharing this framework with your network. Who's one finance colleague who would find this valuable? Sharing would mean the world ๐ŸŒโ™ป๏ธ



2 Ways I can help


When you're ready here's how I can help you:

  1. Book a 1:1 coaching session - We'll turn your tech confusion into tech confidence. Start crushing it here
  2. Join the AI & Automation Academy Waitlist - Get early bird access to my up and coming course here.

Subscribe to Framework Friday

1 tech, productivity, or mental framework to your inbox most Fridays. Access to my tech toolkit, 76 tools and how I've used them โ†’ Many of them free. Links to all my 100s of LinkedIn posts โ†’ Updated weekly

Subscribe for free withย just your e-mail:

I respect your privacy. Opt-out any time.