Block vs De-identify & Allow
By Deidentified · 1 Oct 2026 · 1 min read

Barracuda’s new AI Data Security product is another sign that organisations are moving beyond AI policies and training toward technical controls.
It raises an interesting design question:
What should happen when sensitive information is about to be sent to an AI system?
One approach is to block or warn. For some information, such as credentials, source code or highly confidential material, that may be exactly the right response.
Another approach is to de-identify and allow.
Where the value of the information can be retained without the identifiers, removing that sensitive content before it reaches the AI platform can allow staff to keep working while reducing privacy risk.
These approaches are not necessarily competitors. Organisations will likely need both. The more important shift is that AI governance is moving from simply telling staff what not to do, toward building technical controls into the way AI is actually used.
For us at Deidentified, that means asking:
Can we make the data safer before it reaches AI?
See Deidentified in Action
🔊 Turn sound on