Pseudonymisation Swap the names for codes, and keep the key somewhere else.
Short answer. Pseudonymisation replaces direct identifiers, such as a name or an account number, with a code, so the data can no longer be tied to a person without extra information that is kept separately. The GDPR defines it that way in Article 4(5), and its Recital 26 adds that pseudonymised data is still personal data, because it could be attributed to a person by using additional information. That separates it from anonymisation, where the person is not or no longer identifiable. India’s DPDP Act does not use the word, but rule 6 of the DPDP Rules, 2025 lists masking and virtual tokens mapped to personal data among minimum safeguards. In an AI workflow, a document can be tokenised before processing and re-identified afterwards.
01
Pseudonymisation and its neighbours
NIST’s glossary, quoting ISO/IEC 20889, describes pseudonymisation as a de-identification technique that replaces an identifier for a person with a pseudonym in order to hide that person’s identity.
Technique
What changes
Reversible?
Still personal data?
Pseudonymisation
Identifiers become codes; the key is held separately
Yes, with the key
Yes under the GDPR (Recital 26)
Anonymisation
Data altered so the person is not or no longer identifiable
Not by any means reasonably likely to be used
No under the GDPR, which does not apply to anonymous information (Recital 26)
Encryption
Data scrambled with a key
Yes, with the key
Yes: it is the same data in another form
Masking or redaction
Part or all of a value is hidden, for example only the last four digits shown
Redaction removes it for good; masking may leave the original elsewhere
Often yes, if the rest still identifies the person
02
How it works in a document pipeline
A typical pipeline has five steps. Find the identifiers. Replace each with a token. Store the token-to-identity map separately, with tight access. Process the tokenised text, perhaps with an outside AI service. Swap the tokens back in the output where the reader is allowed to see the real names. OWASP lists tokenisation and redaction, using pattern matching to detect and redact confidential content before processing, among its defences against sensitive information disclosure.
Original
Pseudonymised
Meera Nair of Pune owes ₹4,50,000 on the account ending 4471
[PERSON_1] of [CITY_1] owes ₹4,50,000 on the account ending [ACCOUNT_1]
Dr Sharma saw the patient Arjun Rao on 3 October
[PERSON_2] saw the patient [PERSON_3] on 3 October
03
Limits
Names are not the only identifiers. A date of birth, a pin code and a rare condition together can point to one person. GDPR Recital 26 says to consider all the means reasonably likely to be used to identify a person, including the cost and time needed.
Free text is hard to scan. Pattern matching finds Aadhaar-shaped numbers; it misses “the chairman’s nephew”.
The map becomes the crown jewel. Whoever holds it can re-identify everything, so encrypt it, restrict it and log access. The GDPR definition requires the additional information to be kept separately and subject to technical and organisational measures.
Context can still identify. Test the tokenised text with someone who knows the case.
04
What Indian law says
The DPDP Act, 2023 does not use the words pseudonymisation or anonymisation. Rule 6(1)(a) of the DPDP Rules, 2025 lists securing personal data through encryption, obfuscation, masking or the use of virtual tokens mapped to that personal data among the minimum reasonable security safeguards, from 18 months after the Rules were published in November 2025. Whether tokenised data counts as personal data under Indian law is a question for your lawyer. The safe working assumption is that it does, as under GDPR Recital 26. This is general information, not legal advice.
05
Where ANVE stands
ANVE keeps data on the box in the office, which is a different control from pseudonymisation. Its security page (/security) says cloud models exist only if an administrator adds a provider, and that a policy that blocks confidential data from cloud models is not on ANVE yet. An office that does add a cloud model could pseudonymise documents before sending them, and ANVE’s log shows which calls ran in the cloud. Whether ANVE offers built-in pseudonymisation is a question to settle in writing before a paid pilot.
06
Questions people ask
Is pseudonymised data personal data?
Under the GDPR, yes: Recital 26 says pseudonymised data that could be attributed to a person by using additional information is information on an identifiable person. Treat it as personal data in India too, unless your lawyer advises otherwise.
What is the difference between pseudonymisation and anonymisation?
Pseudonymisation can be reversed with the key you keep. Anonymisation means the person is not or no longer identifiable, which has to hold against all means reasonably likely to be used to re-identify them (Recital 26).
Can I pseudonymise before using a cloud AI service?
It reduces what the provider receives, but free text may still identify people, and the mapping table stays with you. Check the provider’s terms as well, and consider a private model for the most sensitive files.
Does pseudonymisation satisfy the DPDP Act?
The Act and the Rules do not mention it by name, but rule 6 lists virtual tokens among security measures. It is one safeguard among several, not a substitute for the Act’s other duties. This is not legal advice.
07
What is not real yet
No sealed ANVE unit exists; every image of the appliance on this site is a labelled render. The software runs today on a development machine. No customers yet, no certifications, no delivery date, and the final price is not announced.
Reservations are ₹999 in India or US$12 elsewhere, refundable in full until you confirm an order in writing (terms). Offices that cannot use cloud AI can apply to be one of ten design partners.
