Text Embeddings Reveal (Almost) As Much As Text. Embeddings of text—where a text string is converted into a fixed-number length array of floating point numbers—are demonstrably reversible: “a multi-step method that iteratively corrects and re-embeds text is able to recover 92% of 32-token text inputs exactly”.
This means that if you’re using a vector database for embeddings of private data you need to treat those embedding vectors with the same level of protection as the original text.
Recent articles
- OpenAI are quietly adopting skills, now available in ChatGPT and Codex CLI - 12th December 2025
- GPT-5.2 - 11th December 2025
- Useful patterns for building HTML tools - 10th December 2025