Privacy and compliance · 2026-09-25
Questions about EVREN retention, access, and deletion
Instead of asking EVREN 'is my data kept', which questions need to be asked separately, what is publicly known today, and where the terms of use can be read.
Short answer
We found no public text describing how long EVREN keeps content, whether it uses content to train models, or how it handles deletion requests. Terms of use exist, but they cannot be read without an API key. So before organisational use, asking the right questions is worth more than filling the answers in with our own guesses.
We recommend using EVREN through LLMTR: you register your own EVREN key once in Settings, and the same code then reaches both EVREN models and the rest of the catalog. The key stays yours, and EVREN's own logging and retention policy still applies unchanged.
Why one question is not enough
"Is my data kept?" has no single answer, because one request can produce several records. The prompt and answer themselves, a per-request status record, account and quota information, the records of a content moderation layer and the backups of all of these can live in separate systems with separate retention periods. A provider saying "we do not store request content" does not answer what happens to security logs or backups.
The same split applies to access and deletion. A deletion request removing a record from the production database and removing it from backups and archives are different jobs; which one is meant is not clear until it is asked in writing.
The questions and what is known today
The table lists the questions to put to EVREN and, for each, what can be found in public sources today. "Not found" is not a negative answer; it means the answer is not public.
| Question | Why it matters | Publicly known |
|---|---|---|
| Are prompt and answer text stored, and for how long? | Requests containing personal data and trade secrets | Not found |
| Is content used for model training? | The chance of data surfacing in other users' answers | Not found |
| Does the content moderation layer keep its own records? | The schema implies such a layer on generative endpoints | The layer exists, its logging policy is unknown |
| How long is the request-status record kept? | The record is tied to an identity | The record exists, its period is unknown |
| How is deletion requested, and does it cover backups? | Requests under data protection law | Not found |
| What happens when the terms change? | Validity of the accepted text | A material change requires re-acceptance |
Where the terms of use are read
EVREN manages its terms of use through the API. The /v1/terms/status endpoint returns the current version and your acceptance state, /v1/terms/text returns the text itself, and acceptance goes through /v1/terms/accept. The is_material field in the status response shows when a material change requires the terms to be accepted again.
Because these endpoints need a key, the terms can be read only after the account that will accept them exists. Some of the questions in the table may be answered in that text; for anyone who can read it, the first step is to look there for the retention, logging, training and deletion clauses.
The answers on the LLMTR side
We can ask ourselves the same questions. LLMTR does not write request and response bodies to its database; the usage log keeps the model, token counts, message count, cost, status code and duration. Because content is never written, retention, training use and content deletion do not arise on the LLMTR side.
Your EVREN key is stored encrypted and never shown again, even to you; only its last four characters are visible. It becomes unusable the moment you remove it in Settings. Personal data about your account is covered by the privacy policy and the data protection notice.
Until the answers arrive
Until the questions have written answers, the reasonable assumption is that the provider may keep logs; the data notice on the model cards says the same. The assumption does not rule EVREN out, but it decides what gets sent: tests run on sample data, and personal data is removed or pseudonymised before it is sent.
In organisational use it also matters where the answers are kept. Once a written answer from the provider is added to the organisation's data inventory and risk register, the next audit's question of who accepted this, and on what basis, is already answered. When a change to the terms triggers is_material and re-acceptance, that record is reviewed too.
Frequently asked questions
Should I use EVREN directly or through LLMTR?
We recommend using EVREN through LLMTR. Your requests still go to EVREN on your own EVREN key and LLMTR charges nothing for them; the difference is that the same code also reaches the other models in the catalog. EVREN's logging and retention policy is not affected by the choice.
Does EVREN use data to train models?
We found no public text that says so or denies it. The terms of use can be read with an API key from the /v1/terms/text endpoint; the answer may be there.
Does the request-status endpoint show the retention policy?
No. It shows a request's execution and billing state. It says nothing about how long content is kept or where.
Can I accept EVREN's terms of use through LLMTR?
No. The terms are accepted on your EVREN account, from the EVREN console or through EVREN's own endpoints. When EVREN refuses the key the response is 403 provider_credential_rejected, and that error also covers terms that have not been accepted; so on this error the terms acceptance is checked alongside the key.
What happens if I remove my EVREN key from LLMTR?
The key becomes unusable immediately and requests to EVREN rows return 403 provider_credential_required. Revoking the key on the EVREN side is done from your EVREN account.