This article explains how Canto handles customer data across all AI features: what the data commitments are, what Canto does not do with your content, and how third-party AI services are governed.
For the full document, see the Canto AI Policy.
The short answer
Do you use my content to train your AI?
No. Canto does not use customer assets to train or improve the foundational AI models available to other customers. Any analysis performed by Canto's AI is ephemeral, scoped to your tenant, and processed solely to generate the output you requested.
Canto's approach to AI: assistance, not replacement
Canto builds AI to augment human creativity and efficiency, not to replace the judgment of creative professionals or the strategic decisions of its customers. AI features help you search, tag, and organise content more intelligently — every output is a suggestion that a human reviews, edits, or overrides.
AI exists to serve your workflow within your environment. It does not pool data across customers, build profiles of organisations from their content, or use what it processes to improve results for other customers.
Four core principles
Canto's AI development is anchored in four commitments, drawn from the AI Policy:
| Principle | What it means in practice |
|---|---|
| Fairness | AI models are tested regularly to ensure results are equitable across different content types and demographics. Bias and accuracy monitoring is an ongoing process, not a one-time check. |
| Transparency | Where AI informs a decision or automates a step in your workflow, Canto provides clear information about what it did and its limitations. |
| Privacy-by-Design | Privacy is built into AI features from the outset — not added later. Customer content is processed strictly in accordance with your instructions. Pre-trained models are trained on de-identified, lawfully sourced datasets, not on customer assets. |
| Accountability | Canto maintains governance structures to oversee AI development, conducts assessments of new AI features before launch, and operates a formalized incident response protocol for AI-related issues. |
Data integrity and what Canto doesn't do
The following commitments apply to all of Canto's AI features:
- Customer assets are not used to train pre-trained models. The models Canto uses to power AI features are trained on de-identified, lawfully sourced datasets — not on the images, videos, or files you upload to the platform.
- AI analysis is ephemeral. When Canto's AI analyses an asset, it applies what the model already knows. It does not retain a learned representation of your asset after the analysis is complete.
- Analysis is tenant-specific. Results and inferences are scoped to your organisation's data only. Customer data is not pooled, commingled, or shared across tenants.
- Intellectual property stays with the customer. Customer assets remain your intellectual property. Canto's role is to help you unlock value from that content, not to derive any rights or benefits from it.
- Canto acts as processor, not controller. Canto processes data strictly on your instructions. You remain the data controller and determine the purpose and means of processing.
Human-in-the-loop: you stay in control
All AI outputs within Canto are suggestions, not final decisions. You can review, modify, override, or reject any AI-generated content before it becomes part of the permanent record.
No AI output is written to an asset's permanent metadata without a human reviewing and confirming it. Editorial judgment, brand voice, and contextual accuracy remain in your hands.
In the rare event of a significant AI issue or model error, Canto operates a formalised incident response protocol that activates a cross-functional team to investigate, contain, and resolve it.
|
Please note: Under applicable law (including the EU AI Act), customers who use AI-generated content have an obligation to disclose it where required. Where AI output is used in customer-facing materials, you are responsible for ensuring any required disclosure is made. |
AI features overview
Canto currently offers the following AI-powered features:
| Feature | What it does |
|---|---|
| Smart Tags | Automatic image and video tagging, face recognition, and OCR |
| AI Visual Search | Natural language and reverse image search |
| AI Library Assistant | Asset categorisation and assisted metadata tagging |
| AI Bulk Update | Product record search and batch updates (Product Hub) |
| AI Quick Edit | In-app generative image editing |
| Amazon AI Export | Automated product data submission to Amazon (US only) |
How third-party AI sub-processors are governed
Where Canto uses a third-party AI service to deliver a feature, the following standards apply:
- All AI sub-processors undergo a security review, privacy assessment, and contractual review — including execution of a Data Processing Agreement — before being engaged.
- Enterprise-level agreements with all AI sub-processors explicitly prohibit the use of customer data for model training, model fine-tuning, or any secondary purpose.
- Customer data submitted to a third-party AI service is processed solely to generate the requested output and is not retained beyond what the contract permits.
- No cross-customer commingling of data occurs at any third-party AI sub-processor.
- Canto holds third-party AI services to the same security and privacy standards it maintains internally.
The full list of Canto's AI sub-processors is maintained in the Trust Center at trust.canto.com.
Frequently asked questions
Does Canto use my assets to train its AI models?
No. Canto's pre-trained AI models are trained on de-identified, lawfully sourced datasets, not on customer assets. Your content is used only to generate outputs within your own tenant environment — it is not used to improve models that other customers benefit from.
Can Canto's AI see content from other customers' libraries?
No. AI analysis is scoped to your organization's tenant only. Customer data is logically separated and never pooled or shared across tenants for any AI processing purpose.
Do I have to use Canto's AI features?
No. AI features are optional and must be enabled for your account.
You can contact your Account Manager or submit a request via the Help Center to disable any AI feature.
Where can I find the full AI Policy?
The full AI Policy is available at canto.com/legal/cantoai.