Snapshot 299629

Document
Subprocessor list
URL
https://harriethq.com/sub-processors/
Fetched
HTTP status
200
Content type
text/html; charset=utf-8
Fetch mode
static
Size
28025 bytes
SHA-256 (raw)
5c031fdb447cb5881506c87c8d1ff646554bb41f4acda14e8b8505610ed6c08e
SHA-256 (normalized text)
e389501e67e194e64b7a20e717a6ce919efb215db91700b4b85ef3bb06d6956e

Raw (normalized text)

Scripts and page chrome removed; this is what change detection compares.

Sub-processors
Version 1.2 · Effective 17 November 2026 · Last updated 17 November 2026
Harriet (JustParent Ltd, company number 14338607) uses the following sub-processors to provide the Harriet platform, website, and related services.
We give customers at least 30 days' notice before adding or replacing a sub-processor (see our DPA). To subscribe to change notifications, email legal@harriethq.com.
Related documents: DPA · Privacy policy · Terms of use
Platform and infrastructure
Sub-processor	Function	Location	Transfer mechanism
Salesforce (Heroku)	Application hosting and infrastructure	United States / EU regions	UK IDTA / EU SCCs + UK Addendum
Amazon Web Services	Cloud storage (S3), email (SES), and related infrastructure (excluding AI inference; see AI and automation)	EU (Ireland) and other regions as configured	UK IDTA / EU SCCs + UK Addendum; adequacy where applicable
Cloudflare	Network protection and content delivery for Harriet domains, website hosting, and hosting for Sites published from Harriet, including their stored data	Global edge network; stored Sites data in the EU	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
Stackhero	Message queue for background jobs	United States / EU regions	UK IDTA / EU SCCs + UK Addendum; adequacy where applicable
Modal Labs	Isolated sandboxes that run hosted connectors and cloud workspaces, including content from connected apps while a session runs	United States / EU (compute pinned to the EU for customers who require it; storage in the United States)	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
AI and automation
Sub-processor	Function	Location	Transfer mechanism
Amazon Web Services (Bedrock)	LLM API for chat and workflow features via AWS Bedrock, including EU-hosted Anthropic models where configured (sub-processor; Customer Data shared to generate responses; not used for model training; limited retention per provider terms)	EU (Ireland) / United States	UK IDTA / EU SCCs + UK Addendum; adequacy where applicable
OpenAI	LLM API for chat and workflow features (sub-processor; Customer Data shared to generate responses; not used for model training; limited retention per provider terms)	United States	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
Anthropic	LLM API for chat and workflow features (sub-processor; Customer Data shared to generate responses; not used for model training; limited retention per provider terms)	United States	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
Google (Cloud / Gemini)	LLM API and related cloud services (sub-processor; Customer Data shared to generate responses; not used for model training; limited retention per provider terms)	United States / EU	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
Microsoft (Azure OpenAI)	LLM API for EU-hosted GPT models (sub-processor; Customer Data shared to generate responses; not used for model training; limited retention per provider terms)	EU	UK IDTA / EU SCCs + UK Addendum; adequacy where applicable
Baseten	Hosting for open-weight models (DeepSeek, Kimi, GLM) through its Model APIs (sub-processor; Customer Data shared to generate responses; not used for model training; zero data retention)	United States	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
TensorX	Hosting for open-weight models in the EU (sub-processor; Customer Data shared to generate responses; not used for model training; zero data retention)	EU	UK IDTA / EU SCCs + UK Addendum; adequacy where applicable
Groq	LLM API for optional open-weight chat models (sub-processor; Customer Data shared to generate responses; not used for model training; not retained by default)	United States	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
Perplexity	Web search tool (sub-processor; queries and supporting context shared to generate responses; not used for model training; zero data retention)	United States	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
Microsoft	Optional Teams integration, authentication, and mailbox features	United States / EU	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
AI models and data handling
Current model availability and EU hosting are on our Models page. This table shows the models each provider serves for Harriet and what the provider does with data under the commercial terms Harriet uses. We checked it against each provider's published terms on 26 October 2026. The provider's terms govern.
Provider	Models	Terms	Training on Customer Data	Retention
OpenAI	GPT 6.1, GPT 6, GPT 5.6, GPT 5.5, GPT 5.4, GPT 5.2 and GPT 5 families, GPT 4.1, GPT 4o and o-series reasoning models	Commercial API terms	Not used	Up to 30 days for abuse monitoring; zero data retention available for eligible customers
Anthropic	Claude Fable, Opus, Sonnet and Haiku families, including Fable 5, Opus 5.5, Sonnet 5.5 and Haiku 5.5	Commercial API terms	Not used	Deleted within 30 days, with limited exceptions for policy enforcement and law; zero data retention available for eligible models (Anthropic keeps some newer models, such as Fable 5, outside it)
Amazon Bedrock (EU, Ireland)	Anthropic Claude Opus 5.5, 5, 4.8 and 4.7, Sonnet 5.5 and 5, and Haiku 4.5	AWS terms	Not used	Model providers do not receive prompts or completions; AWS may retain them to detect abuse
Microsoft Azure OpenAI (EU)	GPT 6.1, GPT 6 and GPT 5.6 families, and earlier GPT models	Microsoft product terms	Not used	Not available to OpenAI; Microsoft abuse monitoring applies
Google Gemini API	Gemini 3.1 Pro (preview), Gemini 3.8, 3.7, 3.6, 3.5 and 3 Flash, Gemini 3.1 Flash Lite, and earlier Gemini models	Paid Services terms	Not used	Logged for a limited period to detect abuse
Baseten	DeepSeek V4 Pro and V4.1 Flash, Kimi K3, GLM 5.2 and 5.3	Commercial terms	Not used	Zero data retention
TensorX (EU)	DeepSeek V4.1 Flash, Kimi K3, GLM 5.3 and GLM 5.3 Flash	Commercial terms	Not used	Zero data retention
Groq	Llama 3.3 70B and Mixtral 8x7B (optional chat models)	Commercial terms	Not used	Not retained by default; may be logged for up to 30 days to investigate reliability issues or abuse
Perplexity	Sonar Pro (web search tool)	API terms	Not used	Zero data retention
Communications and support
Sub-processor	Function	Location	Transfer mechanism
Mailgun (Sinch)	Transactional email delivery	EU	UK IDTA / EU SCCs + UK Addendum
Slack Technologies	Optional customer integration and notifications	United States	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
Google	Optional Google Chat integration and notifications	United States / EU	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
Payments and billing
Sub-processor	Function	Location	Transfer mechanism
Stripe	Payment processing and billing	United States / EU	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
Analytics, monitoring, and marketing
Sub-processor	Function	Location	Transfer mechanism
PostHog	Product analytics (consent-gated on website)	EU / United States	UK IDTA / EU SCCs + UK Addendum
Google (Analytics / Tag Manager)	Website analytics and advertising tags	United States	UK IDTA / EU SCCs + UK Addendum; DPF where applicable
Attio	CRM for sales enquiries	United Kingdom / EU / United States	EU SCCs + UK Addendum; adequacy where applicable
Sentry	Error monitoring and application performance	United States / EU	UK IDTA / EU SCCs + UK Addendum
Notes
Customer-enabled integrations (for example Slack, Microsoft Teams, Google Chat, or a specific HRIS) process data only when Customer configures them.
LLM providers are sub-processors: relevant Customer Data is shared with them to deliver AI features. They are contractually required not to use Customer Data for model training. Limited retention may apply for service delivery, abuse prevention, and compliance. See our Privacy policy and DPA.
Customer-chosen routes. Customers can use their own provider API keys, connect their own model server (for example vLLM), or use Harriet's connectors from AI clients of their choice (for example Claude Desktop or Cursor). In those cases the provider or client is the customer's own supplier, not a Harriet sub-processor, and the customer's agreement with it governs how data is handled.
Google user data. Data received from Google APIs follows the Limited Use requirements. See the Google user data section of our privacy policy.
This list covers sub-processors used to deliver the Services and operate our website. It is updated when we add or replace providers.
Contact
Questions about sub-processors: legal@harriethq.com