Frequently asked questions

AI Data Handling & Security FAQ

Last updated: November 16, 2026 Effective: November 16, 2026

AI Models & Hosting

Which AI models do you use?

We use AI in layers, and every layer runs inside our controlled environment. First, we apply your translation memory and glossaries, which hold the content and terminology you've already approved. Next comes neural machine translation (NMT). Our primary large language model (LLM) is Claude, developed by Anthropic. For specialized content, we use two proprietary LLMs: one for healthcare and one for market research.

Where is my data hosted?

All data is hosted in the United States. We access Claude through Amazon Bedrock on AWS in the United States. Our two proprietary LLMs and our Translation Management System (TMS) are hosted on Microsoft Azure in the United States.

Does my content leave your secure environment?

No. Your source files stay in our TMS. Linguists work directly in that system and never move content to outside tools. Our platforms connect only through API calls (direct system-to-system connections), and our IT infrastructure secures them.

Data Separation & Access

How is my data kept separate from other clients' data?

We store each client's data in its own section of our database. Your translation memory, glossaries, and style guides stay apart from other clients' resources, and they belong to you.

Who can access my content?

We restrict access by role and by project. Employees and linguists get access only to the resources their role requires. We grant and remove access through scripted onboarding and offboarding procedures and a standard permissions template. Linguists don't receive files. They log in to our TMS to work on assigned tasks, so nobody emails files or stores them locally.

Can I limit access within my own organization?

Yes. In the Argo Translation Customer Portal (ACP), you can limit access by department or position within your organization.

Model Training

Is my content used to train AI models?

No. We never use client data to retrain large language models. We use your translation memory only to train models built for your organization, never general-purpose engines.

Can I get that in writing?

Yes. We'll provide written confirmation that your content isn't used to train models.

Do people review the AI's output?

Yes. Human linguists post-edit all AI output, following ISO 18587:2017, the standard for post-editing machine translation. We use AI and machine translation only with your approval, and we're open about where we use it.

Retention, Deletion & Ownership

Does the AI model keep my data?

No. We use stateless AI models, which don't store your data after processing.

Can I have my data deleted?

Yes, we'll delete it on request. We store client data securely in Microsoft Azure and our TMS.

Who owns my translation memory?

You do. You own your translation memory and all content generated during your projects, including updates made during translation and review. On request, we'll give you a full export in industry-standard formats such as .tmx (Translation Memory eXchange), at no additional cost.

Security Controls

How is my data protected?

Our TMS is hosted on Microsoft Azure in the United States and is SOC 2 and ISO 27001 compliant. We encrypt data at rest with AES-256. Our team reaches the secure Azure file store through Virtual Desktop Infrastructure (VDI). Two-factor authentication (2FA) protects all systems, and every file exchange stays inside our encrypted environment.

How do you protect against threats?

  • SentinelOne, an extended detection and response (XDR) platform for endpoints and servers, protects our environment. Our managed service provider's engineers monitor it.
  • Proofpoint defends against phishing and spear-phishing.
  • Microsoft Purview provides data loss prevention across Microsoft 365.
  • Our IT group patches machines weekly and blocks USB devices on company machines.

What do your people and processes require?

Every employee and linguist signs a non-disclosure agreement and a confidentiality agreement during onboarding. We rotate passwords every 90 days, and all employees complete annual security training. Our ISO 9001:2015, ISO 17100:2015, and ISO 13485:2016 procedures include protocols for handling customer data securely. Read our privacy policy for more detail.

Portal Access

Does ACP support Single Sign-On (SSO) and Multi-Factor Authentication (MFA)?

Yes, both are available as optional add-ons. With SSO, your team logs in to ACP with the Google or Microsoft credentials they already use. MFA adds a second verification step at login, so only authorized users can reach your translation projects, billing data, and reports. Pricing depends on your setup. We'll ask a few questions about your environment, then send a quote.

Need more detail for a security review? Contact your account manager or use our contact form, and we'll walk you through our data handling practices.