Personal data is information that identifies a specific individual directly or when combined with other information. Before sending a dataset to an AI service, ask whether it contains anything that could identify people or expose private details. section 8.6.1
Treat personal data as potentially sensitive and check these safeguards before using generative AI:
- Follow privacy regulations, company policies, and guidance from the organization responsible for the data. section 5.3.3 section 8.6.1
- Apply anonymization or another approved treatment when appropriate; do not assume that an apparently ordinary file is safe. section 8.6.1
- If analysis proceeds, use manageable portions, supervise each stage, and inspect the data and results for errors. section 5.3.3
Checking a data file before upload
A frequency-analysis task may involve uploading a CSV or Excel file to ChatGPT. First check whether names, contact details, or other identifying information appear in the file. If they do, stop and confirm whether anonymization and organizational approval are required; do not rely on the absence of the dataset’s usual records to establish safety, because related proprietary information can still be sensitive. section 5.3.3 section 8.6.1
