Keeping Your Business Data Private During AI Conversations

As more small and medium-sized businesses in Ireland explore AI tools like ChatGPT, it’s natural to have questions about data security. Protecting your sensitive company information during these interactions is essential, and fortunately, it's quite manageable with a few key approaches.

Why this matters for SMEs

Protecting data isn't just about compliance; it's about maintaining trust and avoiding costly breaches. This directly impacts IT managers, operations leads, and indeed every employee who interacts with AI tools, ensuring commercial confidentiality and data privacy.

What it looks like in practice

Consider Fiona, the Marketing Manager at 'Celtic Crafts', a 40-person online retailer based in Cork. Fiona often uses ChatGPT to draft initial social media posts or brainstorm content ideas. Previously, she might have pasted customer testimonials directly into the AI to help generate new ad copy, without thinking twice about where that data was going.

Now, Fiona has a new routine. When drafting a social media campaign for a new product, she needs to reference product specifications and some early customer feedback. Instead of inputting the raw feedback, she first generalises and anonymises it. For example, instead of 'Mary O'Connell from Dublin said the 'Emerald Isle' necklace arrived chipped and was too small,' she'll input 'A customer in the Leinster region found the initial product delivery damaged, and the sizing inaccurate for the 'Emerald Isle' necklace.' She also uses an enterprise version of the AI tool provided by her company, which has been configured not to use her inputs for its own training. This simple shift ensures that no proprietary product details or personally identifiable customer information leaves Celtic Crafts' control, even with an AI assistant.

What could go wrong

• Accidental sharing: Proprietary business strategies or client lists could be inadvertently included in prompts, becoming part of the AI model's training data if not using an enterprise-grade solution.
• Compliance issues: Failing to protect customer or employee personal data when using AI could breach GDPR regulations, leading to fines and reputational damage.
• Competitive disadvantage: Sensitive design ideas or unreleased product specifications could be exposed, potentially benefiting competitors.
• Loss of intellectual property: Original content or unique business processes entered into public AI tools might become less exclusive.

What to try this week

1. Check your AI tool's data policy: For any AI tool your business uses, or is considering, take five minutes to find and read their data usage and privacy policy. Understand how they handle the information you input, specifically whether it's used for model training.
2. Use anonymised or generalised data: Before inputting any sensitive information into a public AI tool, spend a moment to rephrase it. Remove specific names, figures, locations, or details that could identify individuals or proprietary company information.
3. Explore enterprise options: If your team regularly uses AI, investigate enterprise versions of popular tools. These often include dedicated privacy features that ensure your data isn't used for training purposes, providing a more secure environment for commercial use for a relatively small recurring cost.

Being mindful of what you share with AI tools will safeguard your business's valuable information.