liveBook · Manning

Embedding

As explained in AI Governance

An embedding is a numerical vector representation of data that captures features or relationships for computational use.

An embedding is a compact numerical fingerprint of data, such as a sentence. In retrieval-augmented generation (RAG), embeddings are stored in a vector store so a retrieval pipeline can find passages similar to a user’s query. section 5.3.2

Treat embeddings as a potential privacy risk, not as automatically anonymous data:

  • Attackers may reconstruct snippets of original text or infer sensitive attributes, such as a patient diagnosis. section 5.3.2
  • Sort embeddings by source and linkability; those derived from sensitive records should be treated as personal data. section 5.3.2

How AI Governance covers it

How deep it goes
Explains embedding generation, privacy attacks, linkability, and legal scope without detailing embedding internals.
Examples
Demonstrates reconstruction and attribute inference, including inferring a patient diagnosis from embeddings.
What you can do after
Evaluate embedding privacy risk by checking source sensitivity and whether vectors can be linked to people.
How it connects
Connects embeddings to retrieval-augmented generation, vector stores, and personal-data protections under GDPR.

Also explained in

Compare how 72 books explain embedding →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of AI Governance. The section links point to the parts of the book it draws on.