Healthcare edition
This week: how to keep sensitive client data safely out of the chatbot, and why AI fraud hits healthcare particularly hard.
A new OpenAI option: prompts are no longer stored
OpenAI offers 'Zero Data Retention': prompts and responses are deleted immediately after processing.
On 19 August 2026, OpenAI announced 'Zero Data Retention' for certain business customers. This means that what you type in and what the model replies is not stored after processing. OpenAI staff cannot view it. Company data is also not used to train the model, unless an organisation explicitly chooses that itself. The measure is aimed at organisations that work with sensitive data.
Background for subscribersOpenAI has an option that deletes prompts and responses immediately after processing. That sounds safe, but you cannot switch it on yourself. Your organisation must have arranged this as a business agreement. Ask your IT department or administrator whether your licence covers this. If you do not, you cannot be sure what happens to sensitive text you enter. At the same time, this edition warns about AI fraud: cloned voices and fake videos are being used in healthcare too, to force through payments. Always verify an unexpected instruction through a separate, known channel.
Fake boss calls to approve a payment: AI makes the voice almost convincing
In 2026, criminals are increasingly using deepfake videos and cloned voices to impersonate a director or financial officer. This tricks employees into carrying out fake payment instructions. Video and voice verification have as a result become unreliable.
Background for subscribersThis is a taster.
Subscribers read the whole Healthcare edition: every story, each with its background in plain language.
Subscribe →