Otto AI FAQs

The following are common questions and answers about Ask Otto, our generative AI script assistant powered by Automox Otto AI™.

General

Question Answer
Who has access to Ask Otto? The Ask Otto module is included with the purchase of Premium Support or Premium Support Plus services. See Automox Pricing. You must have full administrator or Manage: Organization permissions to enable Ask Otto for an organization. Any full administrator, organization operator, or custom role with Worklets: Create, Execute permissions can use Ask Otto.
Can I disable Ask Otto after it is enabled? To disable Ask Otto after it is enabled, contact Automox Support.
Is Otto AI expected to be able to evaluate the code in an existing worklet without having to put the actual code into the prompt? No, for security reasons we don’t automatically send worklet contents through Otto AI.
Is Otto AI using data from Automox to help formulate its response? No. Otto AI uses Anthropic's Claude Haiku 4.5 model via AWS Bedrock. Claude's responses come from Anthropic's training data — not from Automox customer data. Per AWS's commercial data privacy commitment, Bedrock does not store your prompts or completions, and Anthropic does not use them to train future models.

Security

Question Answer
What is the architecture for Otto AI? The Otto AI backend is a serverless service running entirely inside Automox’s AWS account (us-west-2). Chat requests flow:
  • User → Automox Console (HTTPS) → AWS API Gateway (authenticated) → Lambda → AWS Bedrock → Anthropic Claude (Haiku 4.5).
  • User input never traverses the public internet to a third-party AI provider. The Bedrock API call is internal to AWS.
  • Cross-region inference (AWS-managed) may route the model invocation to us-east-1, us-east-2, or us-west-2; all are within AWS.
What happens to data entered into Otto AI? Is there anything else or anywhere else data sharing is happening? Your prompts and Otto AI's responses are processed by AWS Bedrock (Anthropic Claude Haiku 4.5). They are:
  • NOT sent to OpenAI, ChatGPT, or any public AI service.
  • NOT stored by AWS Bedrock beyond the request lifecycle.
  • NOT used by Anthropic to train future models (per AWS commercial terms).
  • Stored briefly in an Automox-owned DynamoDB table to maintain chat session history. Sessions are automatically deleted 30 days after their last activity.
  • Aggregated, anonymized usage metrics (tokens consumed, request counts) are kept for billing and capacity planning.
How do I know that the code generated by Otto AI is good and not malicious? Otto AI applies multiple safety layers before returning a response:
  • AWS Bedrock Guardrails filter inputs and outputs for hate, insults, sexual content, misconduct, and high-risk content categories. Prompt injection / prompt attack detection is active on user input.
  • PII filtering automatically anonymizes email addresses, IP addresses, and phone numbers; secrets (credentials, tokens) are blocked outright.
  • Topic restriction (in the system prompt) keeps Otto AI focused on endpoint-management worklet scripting. Off-topic requests (creative writing, news, personal advice, non-endpoint software development) are declined.

Even with these layers, Otto AI is a code-generation assistant — always review generated scripts before running them on your fleet, and test in a non-production worklet first. This is true of any AI-generated code regardless of provider.

What if our AI provider has a security incident?

 Otto AI does not depend on OpenAI / ChatGPT. AWS Bedrock and Anthropic's commercial offering have stronger data-handling commitments than the OpenAI consumer ChatGPT product:
  • AWS Bedrock does not persist prompts or completions.
  • Anthropic does not train on customer-submitted data.
  • Bedrock runs inside Automox's AWS account, behind AWS IAM authentication — there is no exposed public endpoint that an attacker could query with our credentials.

A hypothetical AWS Bedrock service incident would be governed by AWS's standard SLA and our enterprise agreement. Automox would be notified through AWS's standard channels.

What happens if an Otto AI generated script breaks something in my environment?

Output from Otto AI should be validated prior to execution on a device. Refer to the agreement that all Otto AI users must accept prior to usage:

The Ask Otto interface allows your users to access AI-generated Worklet script suggestions through the Automox console. Use of Ask Otto will comply with the Anthropic Usage Policy .

Input to Ask Otto is not confidential. Do not include any personal data or proprietary information. Output from Ask Otto is Third Party Content for purposes of your agreement with Automox, and it may be subject to third-party licenses, including open source licenses.

The Output is available "AS IS" without warranty or support by Automox or its AI provider. If you do not want to enable Ask Otto for your organization, please select "Close."
What AI model powers Otto AI? Anthropic's Claude Haiku 4.5 (released October 2025), accessed via AWS Bedrock. We use AWS's cross-region inference profile, which lets Bedrock route invocations to the nearest healthy AWS US region (us-east-1, us-east-2, or us-west-2) for the best latency and availability.
Does Anthropic or AWS train on our data? No. Per AWS's commercial data privacy commitment and Anthropic's published policy for Bedrock customers, your inputs and Otto AI's outputs are not used to train, retrain, or improve any AI model.
What changed when Otto AI moved from ChatGPT to Claude

In 2026, Automox migrated Otto AI from OpenAI's ChatGPT to Anthropic's Claude on AWS Bedrock. The user experience is unchanged — same chat interface, same worklet script generation — but the underlying model, data-handling commitments, and infrastructure improved:

  • Data no longer leaves the AWS ecosystem.
  • Stronger commercial data privacy commitments (no training on customer data).
  • Lower latency (responses are typically faster).

Related Topics