Our products include AI agents that generate content, suggestions, or answers to help you work more efficiently. These tools are an important and increasingly essential part of modern workflows, and many users rely on them. That said, AI is not perfect and may occasionally provide incorrect information.
Our AI agents are always marked with “...”. When using these features, you are interacting with AI systems provided by Criteo.
Key Points:
Before using any Criteo AI agent, keep these points in mind:
AI-generated content is not guaranteed to be accurate or complete.
AI may provide incorrect or misleading information, even when it sounds confident.
Always review and verify important information before doing anything with it.
You are solely responsible for how you use AI outputs. Criteo is not responsible for decisions or actions taken solely based on AI-generated outputs. You further remain responsible for reviewing and verifying such outputs before using.
Use AI as a support and a tool, not your only source of truth.
Criteo AI features are not designed or intended for high-risk uses such as employment decisions, creditworthiness assessments, medical diagnosis, or legal determinations.
Understanding AI Content Generation
AI agents generate responses by identifying and predicting patterns in large datasets. It cannot know or understand facts the way people do. Because of this:
It can sound confident while actually being wrong.
It can miss context or misunderstand your question.
It may guess what you mean if your request is unclear.
It may not reflect recent changes or new information.
It may produce content that is not always consistent with applicable laws or with Criteo advertising guidelines.
AI-generated outputs are produced automatically and are not reviewed, verified, or approved by humans unless explicitly stated otherwise.
What affects AI output?
AI output can change based on how and when you use it:
Results may vary: The same question may produce different answers at different times.
Better input = better results: The more detailed the prompts, selected files, and context you provide, the better output an AI agent will generate.
Accuracy isn’t guaranteed: AI is often helpful, but it’s not always factually correct. It is recommended to verify the sources.
When shouldn’t you depend on AI?
AI outputs should not be used for the following purposes without independent expert review:
Critical decisions where accuracy is essential.
Factual claims without checking reliable sources.
Legal, medical, financial, or safety-related topics.
Real-time or breaking news/information.
Best Practices
To use AI features safely and effectively:
Treat AI output as a first draft. Review, edit, and verify the information before using or publishing it.
Adhere to privacy and confidentiality rules. Do not enter sensitive personal data, trade secrets, or regulated information into a Criteo AI prompt unless your product clearly allows it.
Avoid infringing third-party intellectual property and personal image. Do not use copyrighted materials and personal images for which you don’t have sufficient rights.
Avoid harmful or unethical use. Do not use AI to mislead, harm, discriminate, or create unsafe content.
Follow applicable regulations and rules. Always ensure compliance with applicable laws and regulations, product terms, acceptable use policies, and privacy notices.
FAQs
Why does AI sometimes sound confident, but give wrong information?
AI predicts words likely based on patterns in data. This can result in confident-sounding responses that are incorrect. Always double-check important information.
Can I ask Criteo AI for sources?
Yes, you can ask for sources or references. You should always open the links and check them yourself to confirm accuracy.
Can Criteo access my history of conversations with AI agents?
AI conversations may be stored securely for a limited period and handled in accordance with applicable data protection laws and our Criteo Corporate Privacy Policy. It may be used by authorized Criteo personnel in an aggregated or anonymized form to:
Improve product quality, performance, or accuracy.
Troubleshoot technical issues or investigate errors.
Ensure security, compliance, or prevent misuse of the service.
Is my data used to train Criteo AI agents?
For AI agents developed by Criteo, your input may be used to improve model performance only in specific cases:
When the product explicitly states that user inputs are collected for model training or continuous improvement, and
When you have enabled features or settings that allow data to be used for training, as described in the product configuration options.
If neither applies, your inputs are not used to train or retrain Criteo models. For full details on when and how your data may be used, review your product settings and our Privacy and Data Handling documentation. For AI third-party AI agents integrated into Criteo products, your inputs will not be used to train those third-party models.
What should I do if Criteo AI shows harmful, biased, or unsafe content?
Do not use or share the content. Report it through the agent’s feedback channel or any visible support link so it can be reviewed.
Can I share Criteo AI-generated content directly with customers?
Treat any AI output as a first draft. Review and edit it for accuracy, tone, and policy compliance before sharing it externally. If you still have any other questions about using AI features, data handling, or compliance, or need support, please contact your Criteo representative.