The collections lead sees real values. The model and every vendor see tokens.
Who saw which value, and when. Kept for a year.
ABCPK1234F → [PAN_9f1]The same borrower always gets the same token, so the AI can still follow the case.
dialer ← phone numberA tool or role gets the real value its task needs, and the release is logged.
Aadhaar → not sentConsent covers only the data the purpose needs (Section 6(1)), so extra fields stay away from the model.
| Rule | Requirement | Protecto solution |
|---|---|---|
| Security safeguardsAct s.8(5) Rules 6(1)(a), (b) | Reasonable security safeguards, at the minimum encryption, obfuscation, masking or virtual tokens mapped to the data, plus access control. They cover your processors too. | Protecto replaces personal data with consistent, format-preserving tokens before it reaches a model, index or log. Real values return only to roles your policy names. |
| ProcessorsAct s.8(1), 8(2) Rule 6(1)(f) | You stay responsible for what processors do on your behalf, and need a valid contract that covers security safeguards. | Processors receive tokens, so a model provider or log vendor holds less real data. The audit log shows what each one received. |
| Logs and retentionRules 6(1)(c), 6(1)(e), 8(3) | Logs and monitoring to detect and investigate unauthorized access. Keep personal data, traffic data and logs of processing for at least one year. | Protecto logs every detection and unmask decision. AI logs hold tokens, so a year of retention keeps less real data. |
| Breach noticeAct s.8(6) Rule 7 | Tell affected people and the Board without delay, then send the Board a detailed report within 72 hours. | The audit log shows which processor received which data, and when. That is the starting point for the 72-hour report. |
| Purpose and erasureAct s.6(1), 8(7), 12 | Consent covers only the data the stated purpose needs. Erase data when consent is withdrawn or the purpose ends, and make processors erase it too. | Policy keeps fields the task does not need away from the model. The audit log shows where each value went, so erasure requests have a map. |
| Transfers outside IndiaAct s.16 Rules 13(4), 15 | Transfers are allowed unless the government restricts a country by notification. Significant Data Fiduciaries may have to keep data the government specifies in India. | Protecto runs in India, on your servers or in your cloud. Real values and token mappings stay there, and a model abroad receives tokens. |
| Sector rules on topAct s.16(2) RBI circular, 6 Apr 2018 SEBI cloud framework, 6 Mar 2023 | Payment system data is stored only in India; processing abroad is allowed if the data comes back within 24 hours. SEBI-regulated firms keep cloud data and logs in India. | For those flows, keep Protecto and the model inside India, for example on-premises or air-gapped with a model you host. |
Most of them from May 2027. The DPDP Rules were notified on 13 November 2025 with an 18-month window for notice, consent, security safeguards, breach notice, rights and penalties (Rule 1(4)). Consent Manager rules start one year after notification.
Usually, when it processes personal data on your behalf. Section 8(1) keeps your company responsible for that processing, whatever the contract says, and Section 8(2) requires a valid contract.
Not as a blanket rule. Section 16 lets the government restrict transfers to countries it notifies, and Rule 15 lets it set requirements on making data available to foreign States. Sector rules can be stricter: RBI requires payment system data to be stored only in India, and SEBI's cloud framework keeps regulated data and logs in India.
Rule 6(1)(a) names masking and virtual tokens mapped to personal data as security measures. They sit next to access control, logs, backups and processor contracts. Protecto covers masking, access control by role and logs. Backups and contracts stay with your team.
Tokenization is pseudonymization, not anonymization. Whoever holds the token mapping can reverse a token, so treat tokenized data as personal data unless counsel advises otherwise. With Protecto, the mappings stay in your environment.
Protecto sends tokens instead of real values, which reduces what a model provider receives. That path still carries personal data, so check it against Section 16 and any sector rule. For RBI payment data or SEBI-regulated data, run the model inside India, for example a model you host.
Yes. PAN, Aadhaar and phone numbers are among the 50+ entity types Protecto detects out of the box, and you can add custom entities for identifiers specific to your business. Detection is tested in seven languages, including English. Other languages are available on request and tested on your data.
Protecto is an API product. Your team calls the API before the model, or routes model traffic through Protecto as a proxy, and your app logic stays as it is. Start with one AI flow and one data sample, then add more flows as reviewers sign off.
Basic redaction can, because a blank tells the model nothing. Protecto uses consistent, format-preserving tokens, so the same borrower maps to the same token and the model can still tell who is who.
A practical guide to privacy controls for AI assistants, RAG, and agents. One workflow traced from request to CRM update, with what each destination may see.
By submitting you agree to Protecto’s privacy policy.
A framework for finding where sensitive data moves through your AI systems, before a security review or an auditor finds it for you.
By submitting you agree to Protecto’s privacy policy.