Sub-Processors

Last updated: March 2026

Pressmark uses the following third-party service providers (“sub-processors”) to operate our platform. Each provider processes data under a written Data Processing Agreement that restricts them to processing your data solely on our instructions and in accordance with applicable data protection laws.

To be notified when we add or change a sub-processor, email privacy@getpressmark.com with “Sub-Processor Updates” in the subject line.


AI Service Providers

These providers process your content (voice profiles, research documents, newsletter drafts, audience and industry context) to power Pressmark’s AI features. Pressmark exclusively uses paid commercial API tiers. None of these providers use your data to train their AI models under our commercial agreements.

ProviderServiceData ProcessedData LocationRetentionDPA Status
Anthropic (anthropic.com)Primary AI engine — Claude models for content generation, voice analysis, and AI co-editingVoice profiles, research content, newsletter drafts, conversation contextUnited States7 days (API logs), up to 2 years if flagged by safety classifiersAuto-incorporated via Commercial Terms; includes EU SCCs
Google (ai.google.dev)AI engine — Gemini Flash for real-time AI co-editing and multimodal content processingVoice profiles, research content, newsletter drafts, transcript contentUnited States55 days (paid tier abuse monitoring)Google Cloud DPA applies automatically to paid Gemini API; includes EU SCCs
OpenAI (openai.com)Supplementary AI engine — used for specific generation tasksVoice profiles, research content, newsletter draftsUnited States30 days (abuse monitoring). Note: OpenAI may process fully de-identified, anonymized, and aggregated information derived from your data to improve their systems.Auto-incorporated via Services Agreement (effective Jan 1, 2026); includes EU SCCs

Infrastructure Providers

ProviderServiceData ProcessedData LocationDPA Status
Railway (railway.app)API server, background worker service, PostgreSQL databaseAll user account data, content, AI conversation history, usage logsUnited StatesCloud DPA in effect
Vercel (vercel.com)Frontend application hosting and marketing site hostingServes application assets; does not persistently store user contentUnited States (AWS regions)DPA available via Vercel legal

Authentication & Payments

ProviderServiceData ProcessedData LocationDPA Status
Clerk (clerk.com)User authentication and session managementEmail address, name, password hashes (we never see these), OAuth tokens, session tokensUnited StatesDPA in effect
Stripe (stripe.com)Payment processing and subscription managementPayment card data (we never see or store this), billing address, subscription statusUnited States (PCI-DSS compliant)DPA in effect

Communications

ProviderServiceData ProcessedData LocationDPA Status
Resend (resend.com)Transactional email delivery and marketing audience managementEmail addresses, subscriber metadataUnited StatesDPA in effect

DNS

ProviderServiceData ProcessedData LocationDPA Status
NameCheap (namecheap.com)Domain name registration and DNS managementDomain records only (no user personal data)United StatesStandard terms apply

About Our AI Data Practices

We want to be specific about how your content is handled by our AI providers, because this matters.

No training on your data. All three AI providers contractually commit to not using data submitted through their commercial APIs to train their AI models. This is distinct from their consumer products — Pressmark uses commercial API endpoints with stronger privacy protections.

Temporary retention for safety. AI providers retain API data briefly to detect policy violations (such as harmful content generation). Retention periods range from 7 days (Anthropic) to 55 days (Google). After these windows, your data is automatically deleted from their systems.

One provider may use anonymized data. OpenAI’s agreement permits them to process information derived from your data that has been fully de-identified, anonymized, and aggregated — meaning it can no longer identify you or Pressmark — to improve their systems. This does not include your raw content, identifiable data, or anything that could be traced back to you.

Safety exceptions. If content is flagged by a provider’s safety classifiers as potentially violating their usage policy, it may be retained for longer periods (up to 2 years for Anthropic) for investigation. Additionally, data deletion timelines may be suspended if required by legal processes outside of Pressmark’s or its providers’ direct control.