These figures include demonstration data.

Cyber Behaviour Benchmark

Belgium. This is the live view: it changes as people complete the scan, so it is not for quoting.

The question, exactly as 3,832 people saw it

True or false: if the chat history feature has not been disabled on ChatGPT, your conversations are used to train the tool.
True correct
57%
False
21%
I don't know
22%
The full spread of answers, not just right and wrong. 22% chose "I don't know", which is a different group from the people who confidently picked a wrong answer: one needs teaching, the other needs correcting. Everyone answered in their own language; the English wording is shown here.

What a participant was told

The feedback shown after answering, verbatim:
With chat history enabled, conversations may be stored and used to improve or train the tool. That is why confidential, personal or internal information does not belong in public AI tools.
The workgroup wrote these. Showing them beside the national result is how the questionnaire stays theirs.

Where it is weakest

Share who got it wrong, by group. Only groups with enough people and enough organisations appear.
GroupAnswer wrongPeople
2 to 9 size51%55
Just me size50%32
Other sector49%352
Education sector47%329
Healthcare sector47%402
Technology sector46%428
10 to 49 size46%608
Manufacturing sector44%488
Belgium, everyone43%3,832
250 or more size43%1,886
Public sector sector42%192
50 to 249 size42%1,251
Professional services sector40%486
Insurance sector40%416
Retail sector40%410
Financial services sector37%329
Get a report of your own, and for organisations in Belgium see how yours compares on this. Free, about six minutes per person.See where you stand