NYMETRA for banking and insurance
AI on your client and claims files. With fewer identifiers sent.
NYMETRA pseudonymises identities, IBANs, card numbers and social security numbers on the computer before anything goes to AI.
Claim form
Policyholder: Nadia El Fassi
Social security no.: 2 84 05 69 123 456 78Refund IBAN: BE68 5390 0754 7034
Water damage found in the kitchen, quote attached…
What is at stake
Highly identifying files. A demanding framework.
IBANs, cards, social security numbers, sometimes health data.
The GDPR requires data to be limited to what is necessary. The CNIL restates this principle for social security numbers and health data in insurance.
In a regulated environment, NYMETRA is a complementary measure, to be assessed with your IT and compliance teams.
Use cases
What the AI receives, file by file.
Summarise the claim of [PERS_1] and list the missing documents.
Prepare a reply to [PERS_2] about their financing request.
Write a summary note on the file of [ORG_1].
On quotation
Scoping with your teams.
Environment, access, flows: the configuration is defined with you.
Frequently asked questions
Before you start.
Does a business plan or an API remove the need to minimise or pseudonymise data?
No. A business plan or an API may offer processing terms better suited to professional use, but it does not automatically make every piece of content appropriate to send. You still need to examine which data the task really needs, the applicable confidentiality obligations, the provider’s terms and the account settings. Pseudonymisation is not a universal obligation in every case, but it can reduce exposure when people’s real identities are not needed for the work asked of the AI.
Can SYTEWEB access my documents, prompts or answers?
Not in NYMETRA’s normal operation: original documents, prompts, answers and mapping tables are not sent to SYTEWEB’s infrastructure. A different situation may arise, however, if you request a support intervention that exceptionally requires access to real data. Such access must then be authorised, temporary, limited to what is strictly necessary and governed by the applicable support and data protection rules. No permanent access to client files is provided for.
What happens if NYMETRA misses a piece of sensitive information?
Automated detection cannot guarantee that all sensitive information will be identified in every situation. That is why NYMETRA includes a review step before sending: the user can check, add, remove or correct the items to protect. No request to the AI provider should be considered safe just because an automated system has analysed the document. Human review is part of how NYMETRA is meant to work.
* Internal test: 30 documents filled with public data the engine had never seen; 413 of 429 sensitive items detected automatically (94 to 96% depending on the run). Sources:GDPR, art. 5·CNIL, minimisation in insurance.