Navigating Zapier AI: Legal and Compliance Overview
Date: July 1, 2024
Zapier has a variety of AI products and integrations, from Zapier AI products to help users build automation workflows more effectively to connections with third-party AI apps. Our goal is to help customers take advantage of AI’s benefits in a safe and scalable way.
You can learn more about each of our AI products and how customers use them to solve their business problems on our AI page.
We’re evolving both our policies and this resource based on feedback, so if you don’t see what you need, please reach out to your Account Executive, Customer Success Manager, or our Support team. Contact Support here.
Overview: What is Zapier
Zapier is an online automation platform that our customers use to connect their business applications (apps). Zapier has over 7,000 apps including team collaboration tools like Gmail and Slack, databases and CRMs like HubSpot and Salesforce, and ERP and HRIS systems like BambooHR. Every team can use Zapier, but a significant portion of use cases include Sales, Marketing, Support, Project Management, Data Management, productivity and more.
- Workflow: Zapier’s core product, Workflow, relies on “Zaps”, and enables connections between two or more business applications. How it works:
- Trigger: A Trigger starts the Zap (the connection between the apps) and results in one or more Actions. For example, a Trigger could be receiving a new lead in Facebook Lead Ads.
- Action: An Action could be adding that lead into Salesforce or Hubspot.
- Task: A Task is a successful action taken by a Zap. Tasks are used as a billing unit. Some Zapier tasks are free!
- Tables: Tables is a no-code solution to store, edit, access, and share the data you need to power your Zaps.
- Interfaces: Interfaces allows customers to create custom web pages and apps with interactive components for your users.
AI at Zapier
AI is available for customers to use in two ways: Zapier AI products and third-party AI apps. Zapier AI products and third-party AI apps have slightly different data governance best practices and need-to-knows.
Zapier AI products
Zapier has a variety of AI products and features that aim to help your team by making it easier to brainstorm automation ideas and build your automated workflows faster—regardless of how technical your team is. All you have to do is type what you want in plain language and let Zapier AI features do the rest—no code required. Examples include:
- Zap Builder: Where you can describe a desired workflow in plain (rather than technical) language and get a Zap outline within seconds. This is much faster than building Zaps manually.
- Zapier Copilot: Where you can build alongside an AI chatbot that specializes in Zap creation. Zapier Copilot will guide your team step-by-step to set up app connections and add or change steps.
- AI Fields in Tables: Where you can perform operations on your Tables data by inputting questions or requests in plain language. For example, you could translate the language of your Tables data, summarize your Tables data, or analyze the sentiment of a message.
- Canvas Suggestions / Recommendations: Where you can build a diagram/flow chart, then use Zapier’s AI-enabled feature for suggestions on how to improve what you’ve designed and suggest Zaps that could automate parts of your process that are currently manual to save you time.
See all the AI features and products that Zapier offers here.
Integrating Zapier with AI Apps
In addition to Zapier AI features, you can use Zapier to connect to and automate with over 7,000 apps, including many leading AI apps that you may either already use or be interested in using. These apps will only have access to the information that you choose to send them through Zaps that you build. You are in full control of which apps you want to use for your Zaps.
If you’re on our Enterprise plan, your account admin can turn AI apps off entirely or limit your team to specific ones using App Controls. It’s a governance feature that your account admin can use to add apps to either a list of “restricted apps”, apps that your team cannot use (“block” list), a list of “allowed apps”, or the only apps your team can use (“allow” list). Permissions can get as granular as restricting apps to certain trusted teams and users, or only allowing certain actions.
Our approach to building with AI and model training
Zapier offers AI products to help you build workflows that automate your business. They’ll help you come up with automation ideas, build Zaps, and improve them as your processes evolve. While our AI features have already come a long way, we're dedicated to improving them even further and launching new AI tools for our customers: the more our AI features learn about Zapier-specific concepts and how customers use Zapier, the better they will get for our customers.
We’re committed to building Zapier AI in a principled and transparent way. All of the generative features are powered by third-party out-of-the-box AI models. To make these general-purpose AI models useful for our customers, we customize them by training and fine tuning them.
To customize models, we train them on knowledge about how our Zapier products work, such as how Zaps are formatted, how they can be structured and organized better to improve your automation outcomes, as well as how to build a great Zap and which Triggers and Actions will get the job done for you quickly and efficiently. We also integrate insights about how our customers use Zapier, so we can suggest what works best and get your automations up and running faster and more seamlessly.
Across Zapier’s millions of customers and 12+ years of developing automation technology, we’ve learned a lot about how automation can be used to solve business problems. We’re excited to share this knowledge with you through our AI features. Think of them like having a best friend who happens to know everything about Zapier and is available whenever you need them. The more we model train our AI features, the better they’ll get, and the quicker we can help your team get to ‘process automated.’
Our approach to data usage and opting out
To make Zapier AI useful for Zapier users and specific to automation, we model train Zapier AI with usage information (data about how our customers use Zapier, like structures, triggers, configurations, and feedback).
To improve the accuracy and usefulness of AI for Zapier users, we also model train Zapier AI using customer content (the information, content, or materials that you upload to your Zapier account) after de-identifying it of all personal data.
For any data used for model training, we take careful measures to ensure that no personal data is included and ensure it’s access-controlled and encrypted while at rest and in transit.
No customer content from Enterprise (and Company legacy) plans is used to train AI models; it’s opted out by default. This applies to all Zapier products, including beta releases.
For customers on all other plans, we’ve been providing a simple way for users to opt out since August 2023. Each user that wants to opt out can fill out this form. Once you opt out, it applies to your use of all Zapier products, including beta releases, going forward.
Zapier’s compliance with the EU AI Act
On May 21, 2024, the EU Council unanimously approved the EU AI Act, the world’s first comprehensive regulation of AI technology. We welcome the EU AI Act as an important step toward building trust and fostering responsible development in the use of emerging technologies like AI. Here's our perspective:
- Commitment to compliance: We appreciate the Act's risk-based approach, which distinguishes between different categories of AI systems based on their potential impact. This nuanced approach encourages innovation while ensuring appropriate safeguards are in place for high-risk applications.
- Risk-based approach: We appreciate the Act's risk-based approach, which distinguishes between different categories of AI systems based on their potential impact. This nuanced approach encourages innovation while ensuring appropriate safeguards are in place for high-risk applications.
- Transparency and accountability: The Act's emphasis on transparency and accountability in AI systems aligns with Zapier’s commitment to user trust. We're prepared to implement the necessary measures to ensure our AI features meet these standards.
- Global influence: As a company with a global user base, Zapier recognizes the potential for the EU AI Act to set a standard for AI regulation worldwide. We're proactively considering how to adapt our practices to meet these emerging global standards.
- Fostering innovation: Contrary to concerns that regulations might stifle progress, Zapier believes that clear guidelines can actually spark innovation by creating a level playing field and clear expectations for AI companies, including our teams building Zapier AI.
In short, we’re very much in favor of the EU AI Act. As the specifics of the EU AI Act and other AI regulations become clearer, we aim to work diligently to ensure that Zapier’s AI features are compliant. We see this as an opportunity to further strengthen our commitment to user trust and responsible AI development.
Zapier’s shared responsibility model for AI
At Zapier, we believe in a shared responsibility model for AI. AI safety is a team effort among Zapier, the builder of the third-party AI general purpose foundation model (such as OpenAI, Anthropic, and Google Gemini) that Zapier uses to power our AI products, and our customers as the users of the technology, where everyone plays their part to keep things safe, secure, and responsible.
Your role as a user of Zapier AI products
You control how you use Zapier AI features and are responsible for using them responsibly. Responsible use means following your company’s policies, any applicable laws, and the Zapier Acceptable Use Policy. We also encourage you to educate yourself and your team on AI best practices.
Zapier’s role as the builder of Zapier AI products
We’re responsible for developing, deploying, and operating our AI products in compliance with applicable laws and our own development and security policies and guidelines. We’re also responsible for deeply understanding industry best practices and our customer’s needs, and making sure those policies and guidelines meet them. To date, we’ve implemented appropriate security and data governance measures for the secure and safe processing of your data within Zapier AI products, and we’ll continue to evolve them as our customers' needs and the landscape changes. Lastly, we’re responsible for providing a great product and experience, and therefore, we're continuously improving our AI products so they’re accurate and effective.
AI foundation model providers
We use a variety of AI foundation models to enable AI capabilities in our AI products. You can review them on our Subprocessors page. As the backbone of many AI tools across the web, the builders of these AI foundation models are responsible for complying with applicable laws, their own internal development policies and guidelines for model building, and for implementing measures to limit bias and harmful content. We carefully review the legal and security practices of each subprocessor before using them in our products.
Shared responsibility model for AI
This diagram provides a visual representation of the shared responsibility model:
| Zapier Tech Stack | Shared Responsibility |
|---|---|
Use of Zapier AI Products | By Users: responsible use Each user of Zapier AI products is responsible for their use in compliance with applicable laws and the Zapier Acceptable Use Policy. |
Development of Zapier AI products and Zapier-specific AI models | By Zapier: responsible development and deployment Zapier is responsible for ensuring compliance with applicable laws and internal policies on data use, security, and safety in the development and deployment of the AI application and the Zapier-specific AI models, purpose-built specifically to enhance Zapier automation. |
General purpose AI foundation model | By LLM provider: responsible development and deployment Zapier uses a variety of Large Language Models (LLMs) to power our AI products. The LLM provider is responsible for the development and model building of their general purpose AI foundation model. |
Together, and by each playing our parts, we can work to ensure that your data, and that of your customers, are kept safe and protected.
What's next
This isn’t the end of the road. We will not only continue to lead the charge in AI automation products, but also expand our compliance and legal efforts as AI regulations, best practices, and the AI landscape at large become clearer over time. We encourage our customers to stay informed on AI regulatory developments and actively partner with Zapier in our shared responsibility model.
Zapier is committed to providing a secure, innovative, and customer-centric platform on which our customers can use AI to grow their businesses, build better Zaps, and make technology do more work for them. Please share any questions or feedback whether on AI topics covered by this article or otherwise.
