Privacy Program

Last Updated

June 2, 2026

Mixpanel AI - Trust, Safety, and Compliance

Mixpanel’s AI is built with privacy-by-design principles, giving our customers control over their data to enable compliance.

What are Mixpanel’s AI capabilities?

Mixpanel has woven in AI capabilities across the product with the goal of enabling Mixpanel users to gain faster, high quality insights into how their products and services are performing. Our core AI capabilities include Mixpanel Agent, an AI assistant that helps you complete workflows in Mixpanel (including KPI Agent, Onboarding Agent, RCA Agent, and Dashboard Agent).

Mixpanel also offers a Model Context Protocol (MCP) Server. Once connected, you can bring all your context and insights from Mixpanel to your chosen AI tool, including Claude, ChatGPT, Gemini, etc.

In addition to these core capabilities, Mixpanel has included AI functionality throughout the platform to enhance the analysis and insights. Mixpanel is continuously iterating on our AI offering and bringing new features and insights to customers.

Where can I read more about Mixpanel’s AI capabilities?

What are the potential inputs and outputs for Mixpanel’s AI features?

Inputs to Mixpanel’s AI features may include: natural language queries and prompts, event and user profile data, property names, report configurations, and other Mixpanel project data that a Mixpanel user has permission to access.

These inputs may be processed by our third-party AI providers to render an output.

Outputs may include generated insights, chart configurations, query results, natural language summaries, and suggested actions within the Mixpanel platform.

How does Mixpanel protect and govern data used by Mixpanel’s AI features?

Do Mixpanel’s customer terms apply to AI Features?

Which providers power Mixpanel’s AI capabilities?

Mixpanel utilizes several trusted third-party AI providers, which are listed here.

Mixpanel’s third-party LLM providers will not retain your inputs or outputs because Mixpanel has enrolled in Zero-Data-Retention (ZDR) programs. Mixpanel’s third-party LLM providers are contractually prohibited from using your data to train or improve their models.

Mixpanel’s customers are required to comply with our LLM provider usage guidelines, available here - OpenAI Usage Policies and Anthropic Acceptable Use Policy.

Is my data used by any of Mixpanel’s AI providers (e.g., OpenAI, Anthropic) to train their models?

Do any of the LLM providers store your data?

How does Mixpanel AI use your data?

In addition to the restrictive policies we have put in place for our LLM providers, we also limit the use and access of your data within Mixpanel’s platform. Your data is not used for model training across customers.

Under Mixpanel’s Supplemental Terms, Mixpanel may use your data to train artificial intelligence solely for the benefit of an individual customer. At this time, Mixpanel is not engaged in such model training for individual customers. Nevertheless, we have included this in our Services Agreement because Mixpanel has begun to explore this use for product features on our roadmap.

Under Mixpanel’s Supplemental Terms, Mixpanel may use de-identified inputs and outputs to improve Mixpanel’s AI features. This improvement right expressly does not include training; for example, Mixpanel scores outputs to improve results without model training.

Are Mixpanel’s AI capabilities enabled by default?

Can my organization turn off Mixpanel’s AI capabilities? How?

Yes, you can turn off Mixpanel’s AI features at any time.

Turning off AI features will disable all Mixpanel’s AI features that are powered by third-party AI providers for your organization.

Your Mixpanel admin can make this change by navigating to organization settings, and disabling the “Enable Mixpanel AI” toggle.

Note that once AI-powered features are “turned off” by your Mixpanel admin, everyone in your organization, including owners, admins, and members, will no longer have access to Mixpanel’s AI features.

Am I able to choose which large language models are used for Mixpanel’s AI features?

No. Mixpanel’s AI features use large language models selected by Mixpanel for quality and feature-fit.

Mixpanel does offer a MCP Server that gives you the flexibility to connect Mixpanel to your own instance of a large language model such as Claude, Cursor, Github, ChatGPT, and more. Once connected, your chosen model can work with your Mixpanel data. When you use the MCP Server to connect Mixpanel to your own AI tools, your use of those tools is governed by your agreement with the relevant provider.

Do Mixpanel’s AI features respect the existing permissions and access controls set up in my account?

Are Mixpanel’s AI features considered "high-risk" under the EU AI Act or other emerging AI laws?

No. Mixpanel’s AI features are purpose-built to streamline analytics workflows, enabling users to efficiently construct reports, explore datasets, and derive product insights. These functionalities do not constitute “high-risk” AI systems under the definitions set forth by the EU AI Act. Mixpanel continues to monitor the evolving regulatory landscape across jurisdictions and will update our practices accordingly.

How do you mitigate any bias or discrimination in Mixpanel’s AI features?

Mixpanel takes a multi-layered approach to mitigating bias in AI features. We select third-party large language model providers after thoroughly evaluating their inclusivity practices and bias detection capabilities. We evaluate our chosen LLM providers’ outputs against representative test sets that reflect typical Mixpanel use cases. We treat bias mitigation as a continuous process — monitoring for disparate outputs and updating our approach when issues are identified. AI-generated insights in Mixpanel are designed to inform human decision-making, not replace it. Users remain in control of how insights are interpreted and acted upon.

How do you monitor and evaluate the performance / accuracy of Mixpanel’s AI features?

Mixpanel evaluates our AI features before release and maintains ongoing oversight through continuous monitoring and user feedback. Before release, we define accuracy benchmarks and complete extensive testing against these benchmarks. In-product feedback mechanisms (e.g., thumbs up/down, explicit corrections) feed directly into our quality evaluation process — helping us identify where output would benefit from improvement. When monitoring or feedback identifies a gap, we investigate and continuously iterate.

Our company has a Business Associate Agreement ("BAA") with Mixpanel. Can we still use your AI features?

How should I use these features?

Mixpanel’s artificial intelligence features rely on third-party large language models that generate responses based on your inputs and are probabilistic in nature. Artificial intelligence features can sometimes behave in ways that are inaccurate, incomplete, or unreliable. For example, the responses that you receive may not accurately reflect the content they are based on, or generate content that sounds reasonable but is incomplete and should not be relied on.

We encourage you to review the quality of the responses you receive before relying on them or sharing them with others.

How is Mixpanel ensuring my data is protected when using the AI features?

In addition to the restrictive policies we have in place with our third-party LLM providers, Mixpanel’s AI features follow our existing security practices. Mixpanel has completed the external assessment and compliance certifications for SOC 2 Type II, ISO 27001, ISO 27701. Each request is sent to third-party large language models over an TLS-encrypted service to process and send back to Mixpanel.

Are Mixpanel’s MCP Server third-party connectors enabled by default?

No. Your Mixpanel admin must enable your organization’s access to Mixpanel’s MCP Server. Before connecting to a third-party app, we recommend reviewing the types of data stored in that app, ensuring all user permissions are set appropriately, and confirming that connecting this data aligns with your internal data use policies and practices. You can learn more about Mixpanel’s MCP Server in our documentation.

If I use Mixpanel’s EU or Indian data center, is my data ever sent to the U.S. when I use Mixpanel’s AI capabilities?

Yes, in some cases. Mixpanel’s AI features rely on third-party AI providers whose processing infrastructure may be located outside the EU or India, including in the United States. When you use AI features, your inputs and the corresponding outputs may be transmitted to those providers for the limited purpose of generating the response. The identity and processing locations of these providers are published in Mixpanel’s subprocessor list.

Transfers from the EU and the United Kingdom are made under the European Commission’s Standard Contractual Clauses and, where applicable, the UK International Data Transfer Addendum, and are supported by a transfer impact assessment. Transfers from India are made in accordance with applicable Indian law, including the Digital Personal Data Protection Act, 2023, as in force. The full transfer framework is set out in Mixpanel’s Data Processing Addendum.

Each AI provider is contractually required to: (i) process inputs and outputs solely to generate the requested response; (ii) refrain from using customer data to train its models. All transmissions occur over encrypted connections, and data is encrypted at rest in accordance with industry standards. Mixpanel has implemented Zero-Data-Retention (ZDR) protocols with the LLMs that power Mixpanel AI. No data is retained with LLM providers. Input and output are processed only for the duration required to serve the request.