AI tools are great tools for writing, coding, research, and business. But when providing information to an AI chatbot, extra care should be taken.

Depending on the platform, account type, and privacy settings, the chatbot may save, review, or use your information to enhance its AI. Without having knowledge of their policies, uploading sensitive information could leave you vulnerable to having personal details, client information, financial information, or business information exposed.

Ways to safely use AI tools

1. Adjust platform privacy settings and data controls

Before using an AI platform, check the security and privacy settings to ensure that your conversations are handled in a way that you’re comfortable with.

Some settings you should look for include:

• Turn off model training: If there is an option to opt out of sharing your data or improving the model, make sure this is disabled.
• Use incognito sessions: If you want to avoid saving conversations to your chat history or using them to improve the model, use features like ChatGPT Temporary Chat. This will depend on the retention policies of the platform.
• Manage your activity: Look at the options provided to manage and delete any stored interactions.
• Business/enterprise plans: If you’ll be working with sensitive information, ensure that you choose a service that has clear contractual data protection policies.

2. Remove sensitive information before prompting

Only provide your AI with the information it requires to help you. Remove any personal, financial, or business-related data.

Do the following:

• Remove personal data: This includes removing any names/addresses/phone numbers/email addresses/IDs.
• Never include login details: This includes passwords/bank details/credit card numbers/API keys/private tokens.
• Remove sensitive business data: This includes replacing any sensitive company names/client details/project names.
• Remove any database details: This includes removing database details/server paths, etc.

The purpose of this is to provide enough context for the AI to help you without providing any unnecessary data.

3. Classify information before sharing it

Be aware of how sensitive your information may be before uploading documents, sharing datasets, and prompts.

Use these categories as a guide:

• Public data: These are less likely to pose any risk to your privacy. Examples are published articles, marketing and public documentation.
• Internal company data: This may include processes, reports and regular company documents that should only be shared if allowed by company policy.
• Confidential / Regulated Data: This includes customer data, financial details, health data, legal documents and exclusive information relating to your company. Be aware that this needs a higher level of protection. Only upload this data if you have an approved platform and are able to ensure security and compliance.

This will help identify what type of information is suitable for your AI platform.

4. Choose appropriate AI environments

For tasks involving confidential information, consider AI solutions that provide greater control over data processing, storage, and access.

Options include:

• Local AI models: Running open-source models on private hardware can keep prompts and files within your environment, provided external connections and data-sharing features are properly controlled.
• Enterprise AI services: Consider platforms that make contractual commitments to protect your data, restrict who can access it, and meet recognised security standards. For example, SOC 2 Type II is an independent audit that assesses how well a company maintains controls for protecting customer data over time. ISO 27001 is an international standard that helps organisations establish and maintain a formal information security management system. 
• Internal AI policies: Organizations should establish approved AI tools, define what employees can share, and specify which information requires additional authorization.

These measures help businesses manage AI-related data risks while maintaining productivity.

Conclusion

Reviewing privacy and security settings, limiting who can access shared information, checking whether your data can be used to train AI models, removing sensitive information, and categorising data before uploading it can help reduce the risk of exposure while allowing you to benefit from AI tools. 

Drop your comment below: What steps do you take to protect your personal or business information when using AI tools?

Sources

• A simple way to use AI without leaking private data
• 6 Ways to Use AI Tools Without Exposing Sensitive Data
• Can Your AI Use Sensitive Data Safely? 5 Capabilities You Need