Town uses third-party AI model providers through their APIs
Town says it relies on third-party AI model providers to deliver its product features, and that it reaches them through the providers' APIs. The page presents this as a description of which providers Town relies on and how it protects customer data.
Supporting quotations
The AI model providers Town relies on and how we protect your data.
Town uses third-party AI model providers to deliver product features.
Town uses AI model providers through their APIs.
Qualifications & exceptions
Stated purpose: to deliver product features.
Not stated here
The page does not define 'customer data' or say which kinds of customer content are sent to providers.
Seven listed AI providers may process customer data
The page lists seven providers that 'may process customer data when delivering Town's product features': Anthropic, OpenAI, Amazon Web Services (Bedrock), Google (Gemini), Gemini Enterprise Agent Platform, OpenRouter and ElevenLabs. The word 'may' is the page's own: it lists possible processors, and the entries that follow tie several of them to particular features or routes. Each provider's acceptable use policy is said to be linked where available.
Supporting quotations
AI model providers.
The following AI model providers may process customer data when delivering Town's product features.
Anthropic
OpenAI
Amazon Web Services (Bedrock)
Google (Gemini)
Gemini Enterprise Agent Platform
OpenRouter
ElevenLabs
Qualifications & exceptions
The providers 'may' process customer data; they are not all stated to process it for every customer or feature.
Applies when delivering Town's product features.
Not stated here
The page does not say which data goes to which provider, how long a provider keeps it, or where it is processed.
The page describes Anthropic as Town's primary model provider, used for large language model processing for assistant and routine features.
Supporting quotations
Anthropic
Large language model processing for assistant and routine features. Primary model provider.
Qualifications & exceptions
Described for assistant and routine features.
Not stated here
The page does not define 'assistant and routine features' or say what share of processing Anthropic handles.
Use & Processing · Company practice
OpenAI, Bedrock, Gemini and OpenRouter are used on selected paths
Four listed providers are tied to particular model paths or routes. OpenAI and Gemini are 'used when specific model paths or routing rules select' them. Amazon Web Services (Bedrock) 'applies when Town uses Bedrock-backed model paths'. OpenRouter is 'used only when configured model routes send traffic through' it. Only the OpenRouter entry says 'only'; the other three say when they are used without saying they are used in no other case.
Supporting quotations
Used when specific model paths or routing rules select OpenAI.
Applies when Town uses Bedrock-backed model paths.
Used when specific model paths or routing rules select Gemini.
Used only when configured model routes send traffic through OpenRouter.
Qualifications & exceptions
OpenAI and Gemini: when specific model paths or routing rules select them.
Bedrock: when Town uses Bedrock-backed model paths, for supported Bedrock-backed requests.
OpenRouter: only when configured model routes send traffic through it, for supported LLM requests.
Not stated here
The page does not say which model paths or routing rules select these providers, how often that happens, or whether a customer can see or choose the route.
The page lists Gemini Enterprise Agent Platform (formerly Google Cloud Vertex AI) as managed AI model hosting used for image generation features such as Townie image rendering. This is a separate entry from Google (Gemini).
Supporting quotations
Gemini Enterprise Agent Platform
Managed AI model hosting (formerly Google Cloud Vertex AI). Used for image generation features such as Townie image rendering.
Qualifications & exceptions
Used for image generation features such as Townie image rendering.
Not stated here
'Such as' leaves open whether other features also use this platform.
The page does not say what customer content is sent for image generation.
Use & Processing · Company practice
ElevenLabs handles speech and voice generation when audio features are used
The page lists ElevenLabs for speech synthesis and voice generation, and says its entry applies when audio generation features are used.
Supporting quotations
ElevenLabs
Speech synthesis and voice generation. Applies when audio generation features are used.
Qualifications & exceptions
Applies when audio generation features are used.
Not stated here
The page does not say what inputs (for example text or voice recordings) are sent to ElevenLabs, or whether voice data is retained.
Use & Processing · Company practice
Some providers are used only when specific features are enabled
Under a heading about conditional providers, the page says 'some providers are used only when specific features are enabled' and points the reader to the provider list for when each is used. The statement does not name which providers are the conditional ones; the list entries describe conditions in terms of features or routes.
Supporting quotations
Conditional providers
Some providers are used only when specific features are enabled. Learn more about when providers are leveraged below.
Qualifications & exceptions
Applies to 'some' providers only, and only when specific features are enabled.
Not stated here
The page does not say which providers 'some' refers to, or which features trigger each one.
Inference & Profiling · Company commitment
Town configures providers so inputs and outputs do not train third-party foundation models
Town says it configures each provider and routing layer so that customer inputs and outputs are not used to train third-party foundation models. The fuller statement adds an exception: 'unless a customer explicitly enables or instructs that use'. The introductory paragraph states the commitment without that exception, so the exception belongs with the commitment. The statement is about training third-party foundation models; the page does not address Town's own models.
Supporting quotations
Town configures these providers so customer inputs and outputs are not used to train third-party foundation models.
We configure each provider and routing layer so customer inputs and outputs are not used to train third-party foundation models, unless a customer explicitly enables or instructs that use.
Qualifications & exceptions
Limited to training third-party foundation models.
Unless a customer explicitly enables or instructs that use.
Applies to each provider and routing layer that Town configures.
Not stated here
The page does not say whether Town trains or tunes its own models on customer inputs or outputs.
'Customer inputs and outputs' is not defined.
The page does not address other uses by providers, such as retention, human review or abuse monitoring.
Cross-references
contrast (model-proposed): Routing layers apply controls intended to send data only to aligned providers — The configuration sentence says Town configures each provider and routing layer so inputs and outputs are not used for training; the routing sentence says provider-selection controls are 'intended to' send data only to aligned providers. One is worded as a configuration, the other as an intention; the page does not say how they fit together.
has exception (model-proposed): A customer's explicit enablement or instruction lifts the no-training setup — The no-training configuration applies 'unless a customer explicitly enables or instructs that use'; the exception sits in the same sentence as the general rule, while the introductory paragraph states the rule without it.
A customer's explicit enablement or instruction lifts the no-training setup
The no-training configuration does not apply where a customer explicitly enables or instructs the use of their inputs and outputs for training. The page does not describe how a customer would do that.
Supporting quotations
unless a customer explicitly enables or instructs that use
Qualifications & exceptions
The customer must explicitly enable or instruct the use.
Not stated here
The page does not say how a customer enables or instructs training use, whether this is per account or per feature, or who counts as the customer.
Town's provider agreements prohibit training on customer inputs or outputs
Under a 'No training on your data' heading, Town says its agreements with AI model providers contractually prohibit those providers from using customer inputs or outputs for model training. The sentence speaks of model training only and states no exceptions.
Supporting quotations
No training on your data
Town's agreements with AI model providers contractually prohibit them from using customer inputs or outputs for model training.
Qualifications & exceptions
Covers use of customer inputs or outputs for model training, not other uses.
Not stated here
The page does not say which providers' agreements contain the prohibition or publish the contract terms.
The page does not address provider uses other than model training.
Cross-references
contrast (model-proposed): Town says it verified no-training status with each provider — One sentence says Town's agreements with providers contractually prohibit training; the verification sentence says no-training status was verified through dashboard settings, contractual terms, or API configuration, so for a given provider the basis may be a setting rather than a contract. The page does not reconcile the two or say which providers rely on which.
Inference & Profiling · Company practice
Town says it verified no-training status with each provider
Town says it has verified opt-out or no-training status with each provider through dashboard settings, contractual terms, or API configuration. Because the means are listed as alternatives, a given provider's status may rest on a setting or API configuration rather than on a contract. The broader sentence about provider agreements speaks of contractual prohibitions; the two statements describe the basis for no-training differently.
Supporting quotations
Verified opt-out
Town has verified opt-out or no-training status with each provider through dashboard settings, contractual terms, or API configuration.
Qualifications & exceptions
Verified 'with each provider'.
Through dashboard settings, contractual terms, or API configuration.
Not stated here
The page does not say when verification happened, how often it is repeated, or which provider relies on which method.
No verification evidence is published on this page.
Cross-references
contrast (model-proposed): Town's provider agreements prohibit training on customer inputs or outputs — One sentence says Town's agreements with providers contractually prohibit training; the verification sentence says no-training status was verified through dashboard settings, contractual terms, or API configuration, so for a given provider the basis may be a setting rather than a contract. The page does not reconcile the two or say which providers rely on which.
Sharing & Disclosure · Company practice
Routing layers apply controls intended to send data only to aligned providers
Town says some providers may be configured as fallback or conditional routes, and that where it uses routing layers it applies provider-selection controls 'intended to route customer data only to providers whose data-use commitments align with Town's no-training statement'. The page words this as an intention, not as a guarantee, and does not describe the controls.
Supporting quotations
Routing and conditional providers.
Some providers may be configured as fallback or conditional routes.
Where Town uses routing layers, Town applies provider-selection controls intended to route customer data only to providers whose data-use commitments align with Town's no-training statement.
Qualifications & exceptions
Applies where Town uses routing layers.
The controls are 'intended to' route data to aligned providers.
Some providers may be fallback or conditional routes.
Not stated here
The page does not describe the controls or say how a provider's commitments are judged to 'align'.
The page does not say what happens if a fallback provider's commitments do not align.
Cross-references
contrast (model-proposed): Town configures providers so inputs and outputs do not train third-party foundation models — The configuration sentence says Town configures each provider and routing layer so inputs and outputs are not used for training; the routing sentence says provider-selection controls are 'intended to' send data only to aligned providers. One is worded as a configuration, the other as an intention; the page does not say how they fit together.
Legal Rights & Agreement Rules · Restriction or exception
Customers may not use Town or its AI features for listed prohibited purposes
The page prohibits customers from using Town or its AI features for illegal activity, security abuse, privacy violations, deceptive impersonation, child safety violations, regulated or high-risk decisions without appropriate oversight, harmful content generation, or misuse of voice, likeness, or identity rights. The page does not define these categories.
Supporting quotations
Acceptable use.
Customers may not use Town or its AI features for illegal activity, security abuse, privacy violations, deceptive impersonation, child safety violations, regulated or high-risk decisions without appropriate oversight, harmful content generation, or misuse of voice, likeness, or identity rights.
Qualifications & exceptions
Applies to 'customers'.
The regulated or high-risk decisions item is limited to decisions made 'without appropriate oversight'.
Not stated here
'Customers' is not defined; the page does not say whether it covers individual users as well as organizations.
The page does not define the listed categories or state consequences for breaking the rule.
Town's other terms are not supplied, so other conduct rules are not described here.
Legal Rights & Agreement Rules · Restriction or exception
Regulated or high-risk decisions are restricted when made without appropriate oversight
One prohibited use is regulated or high-risk decisions 'without appropriate oversight'. The wording ties the restriction to decisions that lack appropriate oversight. It does not say what oversight is appropriate or who must provide it, and it does not say whether such decisions with oversight are otherwise permitted.
Supporting quotations
regulated or high-risk decisions without appropriate oversight
Qualifications & exceptions
Limited to regulated or high-risk decisions made without appropriate oversight.
Not stated here
'Regulated', 'high-risk' and 'appropriate oversight' are not defined.
The page does not say who must provide the oversight.
Legal Rights & Agreement Rules · Restriction or exception
Provider acceptable use policies may also apply
The page says provider-specific acceptable use policies 'may also apply' alongside Town's own list, and that each provider's policy is linked where available. The linked policies are not part of the supplied text and are not read here.
Supporting quotations
Each provider's acceptable use policy is linked where available.
Provider-specific acceptable use policies may also apply, as linked above.
Qualifications & exceptions
They 'may' also apply; the page does not say they always do.
Linked 'where available'.
Not stated here
The page does not say which providers have a linked policy, what the policies require, or how a customer is affected when one applies.
Linked provider policies were not supplied and are not summarized.
This page covers AI model providers; backend subprocessors are listed elsewhere
The page says it covers AI model providers only, and that ordinary backend subprocessors are listed separately on Town's subprocessor page. That page was not supplied and is not described here.
Supporting quotations
Scope.
This page covers AI model providers.
Ordinary backend subprocessors are listed separately on Town's
Qualifications & exceptions
Limited to AI model providers.
Covers 'ordinary backend subprocessors' only as a pointer to a separate page.
Not stated here
The subprocessor page was not supplied.
'Ordinary backend subprocessors' is not defined, and the page does not say what they do with customer data.
Customer-connected integrations are governed by the customer's relationship with those services
The page says customer-connected third-party integrations are governed by the customer's relationship with those third-party services and by those services' own terms. This sentence sits in the page's scope section, after the statement that the page covers AI model providers.
Supporting quotations
Customer-connected third-party integrations are governed by the customer's relationship with those third-party services and those services' own terms.
Qualifications & exceptions
Applies to integrations the customer connects.
Not stated here
The page does not list integrations or say what Town sends to them.
The third-party services' terms were not supplied.
Organizations needing more detail on providers or data handling can contact Town
The page invites organizations that need more detail on a listed provider or on Town's data handling practices to contact Town, through a 'Contact us' link to privacy@corp.town.com.
Supporting quotations
Questions about AI providers or data use?
If your organization needs more detail on a listed provider or our data handling practices, we'd love to help.
Contact us
Qualifications & exceptions
Addressed to 'your organization'.
Not stated here
The page states no response time or process.
The page does not say whether individual users, not only organizations, may use this contact.
Legal Rights & Agreement Rules · Context
Page states it was last updated in May 2026
The page says 'Last updated: May 2026'. It gives a month and year only; it states no day, effective date or future version.
Supporting quotations
Last updated: May 2026
Qualifications & exceptions
Month and year only.
Not stated here
The page does not state an effective date, how changes are announced or whether other versions exist.