5 min read
Published on October 16, 2025
As organizations explore the benefits of generative AI, one of the most prevalent concerns revolves around privacy and data governance. At Zoom, we firmly believe that AI should be both powerful and adaptable, conforming to your specific requirements rather than forcing you to adapt to it.
Customers using AI in Zoom Workplace have three distinct deployment options available, where each involves different approaches to data processing and third-party AI models. Customers can reach out to Zoom support to understand their default deployment and to select their preferred AI data processing option based upon their data residency preferences, compliance requirements and account availability.
No matter what deployment option a customer uses for AI features, Zoom's commitment to responsible AI remains. In line with these commitments to responsible AI, we do not use any customer audio, video, chat, screen sharing, attachments, or other communications-like customer content (such as poll results, whiteboard, and reactions) to train Zoom's or its third-party artificial intelligence models. Your content is your content, and this applies to all of our AI deployment options.
If it's important to you to limit data processing to the Zoom Platform, Zoom-hosted Models Only (ZMO) is the option for you. Available AI features are powered solely by Zoom-hosted models running on infrastructure managed directly by Zoom. With ZMO, the AI models are hosted in the region where your account is provisioned, allowing us to process data within the Zoom trust boundary.
Available AI features are powered by Zoom-hosted models running on infrastructure managed by Zoom. That means:
ZMO is a good fit for highly regulated industries or customers with strict internal policies that still want to utilize AI capabilities to enable features like meeting summaries.
However, because this approach does not use external model providers, AI features that rely on third-party model providers are not available with this hosting option. Currently, the ZMO deployment option supports a number of popular features, among them are meeting summaries, chat compose, Whiteboard, content generation, and many others.
For the most up-to-date details on ZMO applicable features, see the "AI Features" section of the Zoom AI Security and Privacy Whitepaper.
For customers who want to use third-party models, but still want to keep AI model processing under Zoom's technical control, Zoom-hosted Models Plus (ZM+) offers a compelling solution.
This approach adds Anthropic's large language models running on Amazon Bedrock to the Zoom-hosted models. Amazon Bedrock is a fully managed service provided through Zoom's existing cloud services provider AWS. Importantly, Bedrock allows AI features to be powered by cutting-edge third-party models—such as those from Anthropic—without sending user content to those model providers.
In short:
While ZM+ provides for an expanded set of AI models, not all AI features are currently supported. For more information, see the "AI Features" section of the Zoom AI Security and Privacy Whitepaper.
For customers who choose to use Zoom's European infrastructure, ZM+ is the default AI deployment option. To change to ZMO, please reach out to Zoom support or your account team. Customers hosted in regions beyond the U.S. or the EU can access available AI features through ZMO. We are currently rolling out ZM+ to additional regions.
For customers looking for the full experience with everything Zoom AI has to offer, utilizing our federated AI approach is the right choice. This approach comes in two flavors:
For customers provisioned on our global data cluster, we currently offer our full portfolio of AI models —from Zoom-hosted models to leading third-party AI model providers and services like Anthropic, OpenAI, and Perplexity.
Our federated AI approach:
This option is perfect for organizations focused on innovation and performance who are comfortable with external model usage and AI data being processed outside our trust boundary. And as with other approaches, inline with Zoom's commitment to responsible AI, we have implemented controls to ensure that customer content data is not used for training models. This includes use of features that require third-party AI providers. For customers provisioned on our local data cluster, for example in Europe, we offer a combination of ZMO and ZM+ with additional models by OpenAI and Elvenlabs available for Zoom Contact Center. The communication with these models is done via API-Calls to data centers in the region.
|
Deployment Option
|
Data Residency Support
|
GDPR Alignment
|
Third-Party AI Model Providers
|
Supported Regional Infrastructure
|
|
ZMO
|
Aligns with data residency preferences for supported regions
|
Supported
|
No
|
US, EU, SG, KSA, AU, IN, CA
|
|
ZM+
|
Aligns with data residency preferences for supported regions
|
Supported
|
Anthropic via Amazon Bedrock (hosted on AWS)
|
US, EU
|
|
Federated
|
Might involve global processing, including the US
|
Supported
|
Anthropic, OpenAI, Perplexity
|
US, EU
|
It's also important to consider other aspects including performance, use cases, and desired feature sets. Whether you're operating in a highly regulated environment or pursuing cutting-edge AI capabilities, Zoom provides the right level of control. For more detailed information on how and why we use data, read our Zoom AI Security and Privacy Whitepaper.
