This topic describes the models provided directly by Automation Anywhere.
Automation Anywhere offers a way to access and use AI models within Automation Anywhere. This approach simplifies the process by providing models directly, eliminating the need for you to manage your own licenses.
- Model selection
-
- When creating a model connection, you select Automation Anywhere as the vendor.
- Here are the models currently available:
Table 1. EU Model Name Model Provider Housting country GPT-4o Azure Sweden, France Claude Sonnet 4.5 AWS Ireland Claude Sonnet 4.5 GCP Belgium Table 2. US Model Name Model Provider Housting country GPT-4o Azure US Claude Sonnet 4.5 AWS US Claude Sonnet 4.5 GCP US - You need to select a region to route traffic, with options for the United States
and Europe.Note: Automation Anywhere provides regional routing options and it is your responsibility to choose the most appropriate region based on your performance and data processing requirements.
- How it works
-
- Automation Anywhere handles the licensing and hosting of the models in the background. There are no additional services or LLM provider accounts that you need to sign up for. Automation Anywhere manages model access and hosting seamlessly in the background, ensuring a simple and unified experience.
- Usage is tracked through a credit system, where each call to the model consumes a certain
number of your Automation Anywhere AI credits.Note: When Automation AI Credits are exhausted, automation and agents that rely on LLM calls will fail to execute those specific model-based steps until additional credits are purchased. Automation Anywhere provides a grace policy that allows continued usage for up to 10% additional credits and 15 days beyond the purchased limit to prevent interruptions while credits are renewed or topped up. For more information, refer to Using Automation credits.
- Each call to an Automation Anywhere provided model is metered based on
the total number of tokens (input + output) consumed. A single model
interaction includes up to 2,000 tokens. When an interaction
exceeds this ceiling, additional model interactions are charged. For
example:
Total tokens consumed Model interactions charged 800 1 2,450 2 9,250 5 - This system is designed to provide a simple, direct connection to powerful AI models for your automations.
- The provided LLM capacity is available exclusively within the Automation 360 platform and can only be utilized when building AI Agents or AI Skills. It cannot be accessed or consumed externally.
- LLMs are deployed as stateless instances, meaning input data is processed but never stored.
- Endpoints are deployed by Automation Anywhere IT within Automation Anywhere cloud environments.
- Instances are regional (US and EU), ensuring tenant traffic remains within the chosen geography.
- Why it matters:
-
- Simplified access: You no longer need to navigate the complex process of securing your own licenses, which saves time and effort.
- Streamlined integration: The models are natively integrated, making it easy to test and use them with existing AI Agents, AI Skills and Task Bots.
- Data control: Regional selection allows you to manage data sovereignty and compliance.
Data & privacy risks evaluated
Automation Anywhere has evaluated key data and privacy risks as part of its deployment of large language models (LLMs). These evaluations ensure you adopt the provided models with confidence in compliance and data handling.
- Customer data being sent to LLMs — does the LLM store the data?
- Data is only processed by the LLM for the purpose of generating the output and is immediately deleted; the LLM stores no data.
- Data residency — where does the data go geographically?
- When setting up the model connection, customer chooses either the EU or US as the deployment region for connecting with the LLM. Accordingly, when customer inputs data, the data does not travel outside the deployment region when traveling to and from the LLM endpoints.
- Compliance with EU & US data residency and transfer guidelines?
- Yes. Automation Anywhere has taken measures, where applicable, to support compliant, data transfers when customer chooses to transfer their data outside its country of origin to the US or EU. Data in transit remains within the chosen, regional boundary..
- Will the LLMs use customer data to train its models or in any other way to improve their products?
- No, for every LLM made available to you, Automation Anywhere has signed agreements prohibiting every LLM from using your data to improve or train their products and algorithms.
- Risk of cross-customer data leakage?
- Evaluation and stress tests confirm no mixing or caching of data between tenants.
- Prompt injection risk — can data leak?
- Evaluations against prompt injection payloads found no data leakage. Stateless design ensures nothing is cached or exposed.