Which AI models each vendor says it runs on

    Model provenance decides who else touches the data and which terms apply downstream. Some vendors name their model providers, some name none. Below is what is on the public record for each vendor we track.

    24 of 39 vendors publish somethingLast checked 25 Sept 2026
    This tracker records only what each vendor has published on its own public pages. It is not an assessment, audit, certification or endorsement. Vendors change their documentation — always verify directly before making a procurement decision.

    What is on the public record

    Which AI models does Abridge use?Ambient scribe
    SummaryChecked 25 Sept 2026
    Vendor presents its Contextual Reasoning Engine as 'leading healthcare AI infrastructure'; the page names no model providers. Separately, the trust centre's Data Service Vendors table lists OpenAI and Anthropic for interim text summarization, Microsoft Azure for backup OpenAI model hosting and Baseten for running Abridge-created models.
    “Introducing Abridge Contextual Reasoning Engine Leading healthcare AI infrastructure”
    www.abridge.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Amazon Connect (Amazon Web Services) use?Contact centre (CCaaS)
    SummaryChecked 25 Sept 2026
    Generative AI-powered post-contact summaries stated to be built on Amazon Bedrock. No individual foundation model or model vendor is named on this page.
    “Because generative AI-powered post-contact summaries is built on Amazon Bedrock, users can take full advantage of the controls implemented in Amazon Bedrock”
    docs.aws.amazon.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Anthropic use?Foundation model
    SummaryChecked 25 Sept 2026
    Anthropic's own Claude model family; model reports incl. training-data description published on Anthropic's Transparency Hub
    “Claude Sonnet 5 was trained on a proprietary mix of publicly available information from the Internet, public and private datasets, and synthetic data generated by other models.”
    www.anthropic.com
    SummaryChecked 25 Sept 2026
    Commercial Terms state Anthropic may not train models on Customer Content from Services
    “Anthropic may not train models on Customer Content from Services.”
    www.anthropic.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Cisco Webex Contact Center use?Contact centre (CCaaS)
    SummaryChecked 25 Sept 2026
    Microsoft named as the third-party model provider for Webex generative AI features; Azure OpenAI named as the provider of content filtering.
    “the 3rd party model provider, Microsoft, does not access, monitor, or store Cisco customer data”
    help.webex.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Cognigy (NiCE Cognigy) use?Voice agent
    SummaryChecked 25 Sept 2026
    Documentation lists selectable LLM providers: Platform-Provided LLM, Azure OpenAI, OpenAI, OpenAI-Compatible LLMs, Anthropic, Google Vertex, Google Gemini, Google GenAI, Aleph Alpha, Amazon Bedrock, Mistral AI, Cognigy Nexus LLM.
    “To add a standard or custom model, select a large language model (LLM) provider integrated with Cognigy:”
    docs.cognigy.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Corti use?Ambient scribe
    SummaryChecked 25 Sept 2026
    Corti states its models are trained only on healthcare data
    “Trained only on healthcare data - no internet noise”
    corti.ai
    SummaryChecked 25 Sept 2026
    Own healthcare model family 'Symphony' offered via API
    “Symphony powers speech, documentation, coding, and reasoning through one API”
    corti.ai

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Coviu use?Platform / infrastructure
    SummaryChecked 25 Sept 2026
    Third-party AI technologies used for transcription and transcript summarisation; providers not named
    “AI technologies provided by third parties for transcription services and for transcript summarization services”
    www.coviu.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Curious Thing use?Voice agent
    SummaryChecked 25 Sept 2026
    ChatGPT named as powering the 'Alex' FAQ voice AI agent, combined with the company's own voice AI technology. No provider is named for the other products.
    “Combining Curious Thing’s voice AI technology and ChatGPT, we bring you a cutting-edge AI agent that can answer business-related frequently asked questions (FAQs) over the phone 24/7.”
    curiousthing.io

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Decagon use?Voice agent
    SummaryChecked 12 Aug 2026
    Vendor states zero-day retention with all AI providers, such as OpenAI and Anthropic, so no conversation data is stored or used for training by them
    “Decagon enforces zero-day retention with all AI providers like OpenAI and Anthropic, ensuring no conversation data is stored or used for training.”
    decagon.ai

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Five9 use?Contact centre (CCaaS)
    SummaryChecked 25 Sept 2026
    Five9 Agent Assist 2.0 with AI Summary stated as powered by OpenAI.
    “Using the same generative AI technology behind ChatGPT, the solution summarizes customer call transcripts in seconds”
    www.five9.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Genesys use?Contact centre (CCaaS)
    SummaryChecked 25 Sept 2026
    Genesys Agent Copilot stated to use a combination of Genesys native models and AWS Bedrock models.
    “Genesys Agent Copilot uses a combination of Genesys native and AWS Bedrock models, depending on the configuration, use case, and dialect.”
    help.genesys.cloud
    SummaryChecked 25 Sept 2026
    Sub-processor list names Microsoft (and separately Google) for speech-to-text / text-to-speech in digital and voice bots
    “Genesys customers using Microsoft for Speech to Text (STT) or Text to Speech (TTS) services. This includes digital and voice bots.”
    help.genesys.cloud

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Google (Gemini / Cloud AI) use?Foundation model
    SummaryChecked 25 Sept 2026
    Google's own Gemini model family; Google states Gemini for Google Cloud does not use prompts or responses to train its models
    “Gemini doesn't use your prompts or its responses as data to train its models.”
    docs.cloud.google.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does HealthEngine use?Voice agent
    SummaryChecked 25 Sept 2026
    Privacy policy names OpenAI, LiveKit, Langfuse, ClickSend and Twilio among third-party service providers supporting its AI virtual receptionist and AI-assisted electronic communication functionality, such as cloud hosting and speech recognition
    “third party service providers who support our AI virtual receptionist and AI assisted electronic communication functionality, such as cloud hosting and speech recognition services, including OpenAI, LiveKit, Langfuse, ClickSend and Twilio”
    healthengine.com.au

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Lyngo AI use?Voice agent
    SummaryChecked 25 Sept 2026
    Anthropic PBC and OpenAI, LLC listed for AI Language Model Services; Deepgram, Inc. for Speech-to-Text Transcription; ElevenLabs, Inc. for Text-to-Speech Voice Synthesis Services.
    “Anthropic PBC Activity: AI Language Model Services Location: United States”
    help.lyngo.ai

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Lyrebird Health use?Ambient scribe
    SummaryChecked 25 Sept 2026
    Azure OpenAI (Microsoft) stated for language processing; Deepgram and Whisper stated for speech-to-text conversion.
    “Azure OpenAI (Microsoft) for language processing Deepgram and Whisper for speech-to-text conversion”
    www.lyrebirdhealth.com
    SummaryChecked 25 Sept 2026
    External AI model providers not named on the compliance page; states notes and transcripts are never used to train external AI models
    “Notes and transcripts are never sold, shared, or used to train external AI models.”
    www.lyrebirdhealth.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Meta (Llama) use?Foundation model
    SummaryChecked 25 Sept 2026
    Meta's own Llama open-model family (Llama 3, Llama 4); Llama 4 model card publishes a training-data overview
    “A mix of publicly available, licensed data and information from Meta’s products and services. This includes publicly shared posts from Instagram and Facebook and people’s interactions with Meta AI.”
    github.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Microsoft Azure AI use?Platform / infrastructure
    SummaryChecked 25 Sept 2026
    Foundry 'models sold by Azure' includes Azure OpenAI models. Azure Speech region tables additionally name gpt-realtime, gpt-4o/4.1/5.x, mai-transcribe, phi4 and Whisper model families as available in specific regions.
    “A 'model sold by Azure' is an AI model designated and deployed as such in Foundry, and includes Azure OpenAI models.”
    learn.microsoft.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Microsoft Dragon Copilot use?Ambient scribe
    SummaryChecked 25 Sept 2026
    Microsoft states Dragon Copilot's AI/ML models are trained solely on anonymized data (customer data anonymized within 90 days in a designated research environment)
    “Dragon Copilot's AI/ML models are trained solely on anonymized data.”
    learn.microsoft.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Mistral AI use?Foundation model
    SummaryChecked 25 Sept 2026
    Own Mistral model family; EU AI Act (GPAI) model documentation hub published on legal.mistral.ai ('Latest model documentations')
    “Welcome to Mistral AI's central hub for documentation and resources relating to the AI Act and other applicable AI Regulations.”
    legal.mistral.ai
    SummaryChecked 25 Sept 2026
    DPA states Mistral AI is authorised to process personal data as controller to train its AI models, unless the customer has opted out
    “Mistral AI is authorized to process the Personal Data as Controller for the purposes of: Training its artificial intelligence models”
    legal.mistral.ai

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Nabla use?Ambient scribe
    SummaryChecked 25 Sept 2026
    Publishes an AI Model Card (CHAI version 2); external model providers not named on trust centre
    “AI Model Card CHAI version 2”
    trust.nabla.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does OpenAI use?Foundation model
    SummaryChecked 13 Aug 2026
    OpenAI states data sent to its API is not used to train or improve its models unless the customer explicitly opts in
    “data sent to the OpenAI API is not used to train or improve OpenAI models (unless you explicitly opt in to share data with us)”
    developers.openai.com
    SummaryChecked 13 Aug 2026
    OpenAI's own GPT model family (current flagship GPT-5.6 line: Sol, Terra, Luna)
    “use GPT-5.6 Sol, our flagship model for complex reasoning and coding”
    developers.openai.com

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does PolyAI use?Voice agent
    SummaryChecked 28 July 2026
    Vendor states it uses Amazon SageMaker to train and fine-tune speech recognition, LLM and TTS models, and Amazon Bedrock to access foundation models. Individual model names/providers not stated.
    “With Amazon Bedrock, PolyAI gains easy access to a range of high-performing foundation models and supporting features around security and safety.”
    poly.ai

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Sierra use?Voice agent
    SummaryChecked 25 Sept 2026
    Vendor states agents are built on a combination of frontier, open-weight and proprietary models. No model provider names stated on the pages checked.
    “Sierra is built on a constellation of LLMs—combining frontier, open-weight, and proprietary models—so your agent can execute seamlessly and reliably.”
    sierra.ai

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Which AI models does Tortus use?Ambient scribe
    SummaryChecked 25 Sept 2026
    OpenAI models generate draft clinical notes, letters and suggested clinical codes from consultation transcripts (transcript text only, not audio); zero-data-retention terms stated
    “OpenAI The consultation transcript text and clinician-provided context (not the audio recording) To generate draft clinical notes, letters and suggested clinical codes from the transcript”
    tortus.ai

    Answered from what the vendor publishes on its own public pages, or from an official register where noted, as at the check date shown. Not an assessment, audit, certification or endorsement.

    Not found on public pages

    We could not find this on the vendor's public pages as at the check date. This is not a statement that the vendor lacks the capability or has not disclosed it elsewhere.

    Right of reply

    Something wrong or out of date?

    Are we missing something, or has this changed? Vendors and their representatives can submit a correction, and we note each change and credit the source. We review every correction request and update the entry when the source supports it.

    Submit a correction