The following frequently asked questions explain AI model usage, data processing, privacy, and model configuration in autonomous agents.

Which model providers do autonomous agents not use by default?

Autonomous agents do not use models from DeepSeek, Qwen, Kimi, GLM, MiniMax, ERNIE, Hunyuan, Doubao, OpenAI, Meta, Mistral, Cohere, or Amazon Nova and Titan by default.

For details on the supported models, see Default LLMs and AI services.

Does Reltio train AI models using customer data?

No. Reltio does not train or fine-tune the LLMs used by autonomous agents with customer data. Reltio also does not create embeddings or vector indexes from customer data as part of the described autonomous agents LLM processing.

For more information, see AI data processing and privacy in autonomous agents.

Where do autonomous agents process AI requests?

Autonomous agents process Reltio-managed LLM requests through Amazon Bedrock using cross-region inference within AWS US regions. When Customer Dataplane is enabled, guardrails operate in the customer's AWS account.

Optional web search requests are processed through Google Cloud Vertex AI or Parallel.

Can I use autonomous agents without Reltio-managed LLMs?

BYO LLM allows your organization to use its own LLM endpoints for supported capabilities. However, guardrails, custom agents, and long-running task eligibility checks continue to use Reltio-managed models.

Can I disable web search in autonomous agents?

Yes. Web search is disabled by default. When web search is not enabled, autonomous agents do not use its optional web search services for agent web searches.

Can the default AI models used by autonomous agents change?

Yes. Reltio periodically updates the default models used by autonomous agents to newer versions. The current defaults are listed in Default LLMs and AI services.