FAQ
Crosstensor frequently asked questions.
Everything you need to know about AI governance, security, technical integration, and Crosstensor pricing.
General questions
What is Crosstensor?
Crosstensor is a governed AI infrastructure for enterprises. It centralizes all interactions between your users, applications, agents, and AI models (LLMs).
It is organized around three complementary building blocks: Toolbox (a ready-to-use AI interface for your business teams), Gateway (a unified technical gateway for managing your LLM flows), and Guard (a cross-cutting security and compliance layer).
Who is Crosstensor for?
Crosstensor is designed for organizations of any size that want to deploy AI at scale while retaining control over access, data, and costs.
Three profiles benefit the most: CIOs and IT architects who want to industrialize AI within their IT systems, CISOs and compliance teams who need to secure and document usage, and SMB/mid-market decision-makers who want to deploy without technical complexity.
What is the difference between Crosstensor Platform and Crosstensor Embed?
Crosstensor Platform is for organizations that want to deploy AI internally. It includes all three building blocks (Toolbox + Gateway + Guard) and lets your employees use governed AI assistants.
Crosstensor Embed is for software vendors and IT departments who want to integrate Crosstensor's AI infrastructure (Gateway + Guard only) into their own applications via an API. The Toolbox is not included.
Do I need to deploy specific infrastructure to use Crosstensor?
No. Crosstensor is offered primarily as a SaaS/PaaS solution: the platform is hosted and operated by Crosstensor, accessible via your browser or an API. You can get started within minutes, with no specific deployment required.
For organizations with strong constraints (regulated sectors, highly sensitive data).
What is the difference between shadow AI and shadow IT?
Shadow IT covers any software or service used without IT approval. Shadow AI is its most recent and riskiest form: consumer AI tools fed with internal data, from personal accounts, with no trace and no control.
Crosstensor answers it with a professional access as simple as those tools, which makes usage visible instead of banning it. Details in the guide adopting AI without losing control.
Do you need an AI usage charter before deploying a tool?
A charter is useful, but waiting for it to be finalised delays access to a professional tool and prolongs shadow AI. A minimal framework is enough to start: authorised data, open models, a named point of contact.
The charter is then consolidated on real usage, as explained in the guide section where to start.
Is Crosstensor an AI Management System (AIMS)?
Yes. An AIMS is the set of rules, roles and evidence that lets an organisation manage AI over time, under the ISO/IEC 42001 standard. Crosstensor is its operational version: access to models, rules applied to every exchange and evidence produced automatically.
Certification of an organisation is still issued by a third-party body. See the guide AIMS, AI Management System.
Security & compliance
How does Crosstensor protect my organization's sensitive data?
Crosstensor Guard analyzes all incoming and outgoing traffic in real time. It detects and can filter or anonymize personal, medical, financial, or confidential data before it is sent to AI models.
You define the security policies suited to your context: blocking, alerting, anonymization, or enhanced logging depending on data sensitivity.
Is Crosstensor GDPR compliant?
Yes. Crosstensor is designed with GDPR compliance as a baseline requirement. The platform includes: full traceability of interactions, granular access rights management, filtering of personal data before it is sent to LLMs, audited logging, and hosting entirely in France.
What is the AI Act and how does Crosstensor help me comply with it?
EU Regulation 2024/1689 (the AI Act) imposes progressive obligations on organizations using AI. The main rules will progressively apply starting in August 2026.
Crosstensor natively generates the required evidence: a registry of AI systems in use, logging of interactions, applied governance policies, and documentation of human oversight mechanisms. This lets you demonstrate compliance without additional development.
What is prompt injection detection?
Prompt injection is an attack that manipulates an AI model by embedding malicious instructions in a request, aiming to make it bypass its safeguards or exfiltrate data.
Crosstensor Guard analyzes every request before it is sent to the model and detects known attack patterns. When detected, the request can be blocked, modified, or flagged according to your security policies.
Can data sent to an AI model be used to train it?
It depends on the provider and the type of offer: the terms of consumer tools and those of professional offers are not the same. Crosstensor accesses models through professional offers whose terms exclude training on your exchanges, and Guard masks sensitive data before sending.
The point is detailed in the FAQ of the AI governance guide.
Where is the data hosted?
The Crosstensor infrastructure is operated in France. Model providers only receive requests after anonymisation by Guard: the mapping table that links pseudonyms to real data stays within that perimeter and is never transmitted to them.
On hosting and the CLOUD Act, see the anonymisation guide; the criteria of a sovereign AI are detailed in the dedicated guide.
Can you send customer data to ChatGPT?
Not as is. A customer file pasted into a consumer tool goes to a third party, often from a personal account and without any trace. With Crosstensor, the same models are available, but the request first goes through Guard, which replaces sensitive data with pseudonyms.
The mechanism is explained step by step in the anonymisation guide.
What is the difference between anonymisation and pseudonymisation?
Anonymisation is irreversible: nobody can find the individuals again, and the data falls outside the scope of the GDPR. Pseudonymisation replaces identifiers with placeholders and keeps separately the key that leads back to them. This second, reversible method is what makes it possible to use AI on real files.
Full comparison in the anonymisation guide.
Can you use AI while respecting professional secrecy?
Yes, provided what was entrusted to you is never revealed to the model provider. By masking identities, amounts and references before sending, a notary, a lawyer, an accountant or a healthcare professional lets AI work on a file without exposing its content.
Use cases by profession are described in the anonymisation guide.
Technical integration
What is MCP (Model Context Protocol) and why does it matter?
MCP (Model Context Protocol) is a standard introduced in 2024 that allows AI models to communicate securely and in a structured way with external data, applications, and services. It has quickly become a de facto standard, adopted by major software vendors (Oracle, Microsoft, Salesforce, HubSpot, Workday…).
Crosstensor includes an MCP client (to access data exposed by other systems) and an MCP server (to expose your own data in a controlled way). This lets your AI models query your ERP, CRM, and internal databases directly.
Which AI models are compatible with Crosstensor?
Crosstensor is compatible with all major LLM providers through its unified API: OpenAI (ChatGPT), Anthropic (Claude), Mistral, Google (Gemini), and any model exposing a compatible API, including models hosted locally or on dedicated infrastructure.
You can switch models or use several simultaneously without changing your integration.
How long does it take to integrate Crosstensor into my IT system?
To get started in SaaS mode with the Toolbox: a few hours. Your employees can access their governed AI assistants from day one.
For an API integration (Gateway) into your existing applications: a few days to a few weeks depending on the complexity of your IT system and the number of MCP connectors to set up. The Crosstensor team supports you at every step.
Is Crosstensor compatible with an agentic architecture (AI agents)?
Yes. Crosstensor is designed for agentic architectures, where a single user trigger generates several successive calls to models. Each call passes through the Gateway and is subject to the same governance and Guard security rules.
Only actual calls to an AI service are billed: an agent's internal orchestration steps are not counted.
Are the real data restored in the answer?
Yes, automatically. The model answers on typed pseudonyms such as [PERSON_1] or [AMOUNT_1]; on the way back, Guard replaces them with the original values before display. You read a complete answer, with nothing to rework.
See automatic restoration in the anonymisation guide.
Does anonymisation degrade the quality of AI answers?
No, if it preserves meaning. Redacting with “XXX” makes the model lose the links between pieces of information; a typed pseudonym keeps the nature of each data point and its place in the document, and the model reasons on the same structure.
A before-and-after example is shown in the anonymisation guide.
Pricing & billing
What is a Crosstensor credit?
A credit is Crosstensor's billing unit. 1 credit = €0.001 excl. VAT. Credits are purchased as a prepaid wallet and consumed as you use the platform.
Each AI interaction consumes two components: a fixed platform fee (governance, Guard, logging) and the compute cost of the models passed on at cost price. You have full visibility into consumption from the back office.
Is there a free trial period?
Yes. We offer a free, personalized demo so you can explore the platform in your specific context. Contact us via the demo request form and our team will respond within 24 hours.
How do you measure the real cost of AI in your organisation?
By consolidating everything that is scattered today: individual subscriptions, licences, credits and API calls, then comparing it with usage and the value created. The Crosstensor financial dashboard does this consolidation per team, per user and per model.
Billing models are compared in the governance guide.
In brief
This page answers the most common questions about Toolbox, Gateway, Guard, pricing, and data security. Each answer is organized by topic so you can quickly find the information you need.