A vector database stores embeddings of your text or images. When someone asks a question, it finds the stored items closest in meaning to that question, not only the ones with the same words. For example, a staff member asks a chatbot how many vacation days new hires get. The vector database pulls up the handbook paragraph about paid time off, even though that paragraph never says “vacation.” It is a common part of retrieval-augmented generation systems, where an AI tool looks things up in your documents before it answers.
Why it matters
A vector database holds copies of your documents’ content, so it needs the same access controls as the documents themselves. Suppose you load donor records or HR files into it. Anyone who can use the tool connected to it may be able to get those details back out, even people who could never open the original files. Before you connect an AI tool to your files, ask the vendor three questions: where is the vector database stored, who can search it, and does it follow the same permissions as your shared drive? To see problems that have been reported in AI tools, check our vulnerabilities page.
Published Updated