Skip to content
PiiBlocker
Back to blog

Is TypeSafe Jev Safe for Customer Data? Retention, Training, and PII

PiiBlocker Team

typesafejevzero data retentionapi privacy

TypeSafe Jev is safe to use with customer data if you take two precautions. Its privacy policy clearly states that TypeSafe will not train or fine-tune any AI model on your input, which is a strong commitment. But the same policy sets no fixed retention period, and it does not mention zero data retention. Your data may be stored for as long as TypeSafe considers reasonably necessary, on servers in the United States. So before sending customer records to Jev, confirm zero data retention with TypeSafe in writing, and strip out personal details the decision doesn't need. This guide covers what Jev is, what its privacy policy promises, what it leaves out, and a practical checklist for using Jev with personal data.

What is TypeSafe Jev?

TypeSafe Jev is a decision model for software, not a chatbot. It launched on September 15, 2026 as the first model from TypeSafe AI, a San Francisco startup, which calls it a "System One" model after the fast, intuitive mode of human thinking. Jev does not write text or code. You send it data plus a set of typed questions, and it returns structured answers with probabilities. It supports three question types: a choice picks one option from a set, a score rates something against defined levels, and a yes/no question returns the probability that a statement is true. TypeSafe reports response times of 70 to 500 milliseconds and input pricing far below general-purpose models. Jev is available through TypeSafe's own API, OpenRouter, and the Vercel and Netlify AI gateways. Developers use it to classify support tickets, sort contact forms, filter spam, and check AI agent actions before they run.

Why Jev sees personal data

Jev sees personal data because the decisions it makes are about messages people write. One AI gateway's launch guide shows Jev routing a contact form submission to sales, support, or spam. That is a sensible use, and it also means every name, email address, and phone number in the form goes to a third-party API. Support tickets carry more. Customers write things like "my card ending 4417 was charged twice" or "I can't log in since my surgery last week." To route that message, the whole message is usually sent. Nothing about this is unique to Jev. Every AI classification service works the same way, and the same caution applies to all of them. What changes from one provider to the next is what happens to your data after the request is processed. That is decided by the provider's privacy policy and your contract with them, which is why the policy is worth reading closely.

What TypeSafe's privacy policy promises

TypeSafe's privacy policy makes two clear commitments about your data, and both are good news. We read the published policy, last updated November 19, 2025. First, TypeSafe will not train or fine-tune any AI or machine learning model on your prompts or other input. This is stated plainly as a default, with no opt-out setting to find and no enterprise plan required. Many AI providers offer no-training only as a toggle or a paid tier, so this is a stronger position than average. Second, TypeSafe will not disclose your input to third parties other than the service providers it uses to run its services. Your data is not sold or shared for advertising. For a company that launched publicly two weeks ago, these are meaningful commitments, and developers evaluating Jev should give TypeSafe credit for them. The concerns below are about what the policy leaves out, not about these promises.

What the policy doesn't say

The policy's gaps matter as much as its promises. First, there is no defined retention period. Personal data is kept for as long as reasonably necessary to provide the services, or in support of TypeSafe's business or commercial purposes, and deleted when you request it. There is no stated number of days and no statement that input is discarded after processing. Second, zero data retention is not mentioned. Several third-party directory sites say TypeSafe offers zero data retention to enterprise customers on request. We could not find that commitment in TypeSafe's published privacy policy, or any zero data retention documentation on TypeSafe's site. It may be available through enterprise contracts, as with many vendors, but it is not a public promise. Third, the services are hosted in the United States. Sending data about people in the EU or UK to Jev is an international transfer that your data processing agreement must cover.

Question Published answer
Is my data used to train models? No
Is my data shared with third parties? Only service providers
How long is my data kept? As long as reasonably necessary; no defined period
Is zero data retention available? Not mentioned in the privacy policy
Where is my data processed? United States

No training is not the same as zero retention

No training and zero retention are two different promises, and for personal data the difference matters. No training means your data does not shape the model, so nothing you send influences the answers other customers receive. Zero retention means your data is not stored at all once your request has been processed. It exists only for the milliseconds the model needs it. Data that is never used for training can still be stored on a provider's servers. Stored data can be exposed in a breach, requested in a legal case, or viewed by staff during an investigation. Only zero retention removes those risks at the provider's end, because there is nothing left to expose. TypeSafe's published policy clearly gives you the first promise. It does not publish the second. We explained the same distinction for another popular AI gateway in our guide to OpenRouter zero data retention, which is worth reading if you route Jev through OpenRouter.

How to use Jev safely with personal data

You can use Jev with personal data safely by treating it like any other processor of customer data. Five steps cover most cases. First, ask about zero data retention in writing: email TypeSafe at [email protected] before sending customer records, and if it is available on your plan, put it in your contract rather than relying on an email. Second, strip what the decision doesn't need. Routing a ticket to billing rarely requires the customer's phone number or card digits, so mask or remove those details first and keep the original in your own system. Third, send the minimum. If a subject line and first paragraph are enough to classify a message, don't send the whole thread and its attachments. Fourth, check that your data processing agreement covers a US-based processor, especially for EU or UK data. Fifth, keep health, financial, and identity data out entirely whenever the decision can be made without it.

The bottom line

TypeSafe Jev is a good choice for fast classification, with a better-than-average training policy and an open-ended retention policy. For everyday data that balance is fine. For customer personal data, confirm zero data retention before you build, and mask what the model doesn't need to see. The safest personal data is data that never leaves your system. That principle is why we built PiiBlocker, a free Chrome extension that masks personal data in your browser before it reaches AI chats. It protects you on seven AI platforms: ChatGPT, Claude, Gemini, Grok, DeepSeek, Kimi, and Perplexity. PiiBlocker works inside the browser, not in API pipelines like Jev, so for developer workflows you will need to mask data in your own code. The rule is the same in both places: mask first, then send. And if a provider's policy doesn't answer your retention question, ask until it does.

Frequently asked questions

Does TypeSafe Jev train on my data? No. TypeSafe's privacy policy states that it will not train or fine-tune any AI or machine learning model on your prompts or other input. This applies by default, with no setting to change.

Does TypeSafe Jev offer zero data retention? TypeSafe's published privacy policy does not mention zero data retention, and we found no zero data retention documentation on its site. Third-party directories say it is available to enterprise customers on request. Confirm directly with TypeSafe at [email protected] before relying on it.

How long does TypeSafe keep data sent to Jev? The privacy policy sets no fixed period. It says personal data is kept as long as reasonably necessary to provide the services or support TypeSafe's business purposes, and deleted on request.

Is Jev GDPR compliant for EU customer data? Jev is hosted in the United States, so sending EU or UK personal data to it is an international transfer. Make sure your data processing agreement covers it, and minimise the personal data you send.

Should I mask personal data before sending it to Jev? Yes, whenever the decision doesn't need it. Routing or classifying a message rarely needs names, phone numbers, or card details. Masking them first reduces your risk regardless of the provider's retention policy.

Sources