Sub-processors and policy changes
We email every owner and admin about a change to this policy or to our sub-processors, and list it at cresva.ai/subprocessors. A change takes effect at least 30 days after we announce it; a correction that only describes what already happens is published and emailed the same day.
Changes, newest first
Correction, announced 2026-11-17
Our privacy policy now lists every sub-processor from the same register the DPA uses. Three were in use and missing or unnamed, and are now listed: Modal, which runs video stitching for creatives you ask for; Resend, now named as the relay for sign-in emails as well as every other email; and Google Workspace, which holds the hello@cresva.ai mailbox.
Pages: /privacy, /dpa, /gdpr, /subprocessors
Change, announced 2026-11-17, takes effect 2026-12-17
A record that a security-relevant action happened (a sign-in, an invitation, a setting changed) will be kept for 400 days from when it happened, even after the account is deleted. On deletion it is unlinked from the company, brand and person, every personal detail in it is removed, and it is deleted at 400 days.
Pages: /privacy, /gdpr, /ccpa, /dpa
Correction, announced 2026-11-17
Our legal pages now state what already happens: where your data is stored, that we share personal information only when marketing cookies are on, which processors are told to delete, that a person handles every deletion request, the autonomy levels by the names the product uses, and section references that point to the right sections. Our postal address is on every page and email.
Pages: /privacy, /terms, /dpa, /gdpr, /ccpa, /data-deletion, /refund, /security
Sub-processors
Hosting and infrastructure
- Vercel: Application hosting, CDN, file storage (Vercel Blob) and web analytics. Data: All application traffic; files you upload or we generate; page views with identifiers removed from URLs. Region: Singapore (application functions, sin1) and the United States (browser extension functions, iad1); pages are served from the edge nearest the visitor. Retention: Request logs per Vercel's plan retention; files until deleted.
- Modal: Compute for stitching generated video creatives, and for media mix model fits when they are switched on. Data: Links to the generated video clips of a creative you asked for; for a model fit, a brand's daily spend and sales series with no customer personal data. Region: Per the vendor's data processing terms. Retention: Per Modal's terms. Only when a creative's video is stitched, or a model fit runs.
- Inngest: Background job orchestration. Data: Job identifiers and ids of the records a job works on. Region: Per the vendor's data processing terms. Retention: Per Inngest's event retention.
- jsDelivr: Public script CDN for charts, previews and code sandboxes the agents generate. Data: The viewer's IP address and browser headers when a generated preview loads a script. No content you create is sent. Region: Global. Retention: Per jsDelivr's terms.
- Tailwind Labs: Stylesheet CDN for landing page mockups the analyzer generates. Data: The viewer's IP address and browser headers when a generated mockup is opened. No content you create is sent. Region: Global. Retention: Per Tailwind Labs' terms.
Data storage
- Neon: Primary PostgreSQL database. Data: All account, brand and connected-platform data. Region: Singapore (AWS ap-southeast-1). Retention: Until deleted by you or by our retention schedule; Neon keeps point-in-time history for 7 days, so deleted data leaves backups within 7 days.
- Cloudflare R2: Storage for generated video creatives. Data: Video, image and audio assets generated for your creatives. Region: Global. Retention: Until deleted.
- Upstash: Cache and rate limiting (Redis). Data: Cached query results, answers and generated pages, keyed by brand. Region: United States. Retention: Each entry expires, between 4 hours and 30 days.
AI services
- OpenAI: Language, vision, embedding and image models. Data: Your questions, the brand data needed to answer them, and generated content. Region: United States. Retention: Chat and response requests are sent with store set to false, which turns off OpenAI's stored completions, and a pseudonymous tenant identifier. OpenAI keeps API inputs and outputs for abuse monitoring for up to 30 days unless zero data retention is agreed; embedding and image requests carry no such setting. Training: Not used to train OpenAI models unless the customer opts in; API data kept up to 30 days for abuse monitoring (developers.openai.com/api/docs/guides/your-data).
- Anthropic: Language models for agents, planning and research. Data: Your questions, the brand data needed to answer them, and agent working context. Region: United States. Retention: Per Anthropic's commercial terms; requests carry a pseudonymous tenant identifier. Training: Not used to train Anthropic models by default (privacy.claude.com, "Is my data used for model training?").
- Google (Gemini API): Image and video generation and AI visibility checks. Data: Generation prompts, reference images and product queries. Region: United States. Retention: Per Google's Gemini API terms. Training: Paid services: not used to improve Google products. Unpaid services: used to improve them (ai.google.dev/gemini-api/terms).
- Perplexity: AI visibility checks. Data: Brand, category and product queries. Region: Per the vendor's data processing terms. Retention: Per Perplexity's API terms. Training: Not retained and not used to train models (docs.perplexity.ai/guides/privacy-security).
- Replicate: Image generation. Data: Image prompts. Region: Per the vendor's data processing terms. Retention: Inputs, outputs and logs are removed by Replicate after one hour. Training: Its terms do not say API data is used for training; inputs, outputs and logs are deleted after one hour.
- Sync Labs: Lip-sync for generated video. Data: Generated video and audio for a creative you requested. Region: Per the vendor's data processing terms. Retention: Per Sync Labs' terms. Training: Its terms do not say; lip-sync is switched off for every brand, so nothing is sent today.
- ElevenLabs: Voice generation for video creatives. Data: Script text for a creative you requested. Region: Per the vendor's data processing terms. Retention: Per ElevenLabs' terms. Training: May be used to improve ElevenLabs models unless the account opts out (help.elevenlabs.io, "Is my data used to improve ElevenLabs AI models?"); voice-over is switched off for every brand, so nothing is sent today.
Observability
- Langfuse: AI request tracing. Data: Model name, token counts, latency and pseudonymous identifiers. Prompt and answer content only if we enable content capture, and then redacted. Region: European Union. Retention: Per our Langfuse project settings. Only when Langfuse keys are configured.
- Sentry: Error tracking. Data: Error details with secrets and personal data redacted, and account identifiers. Region: United States. Retention: Per our Sentry project settings.
- Better Stack: Log storage. Data: Application logs, with secrets redacted and email addresses replaced by a hash. Region: Per the vendor's data processing terms. Retention: Per our Better Stack source settings. Only when a Better Stack source token is configured.
Analytics
- PostHog: Product analytics. Data: Usage events and page paths with identifiers and query strings removed. Region: United States. Retention: Per our PostHog project settings. Only after you consent to analytics.
- Google Analytics, Google Tag Manager and Meta Pixel: Advertising and analytics measurement on public cresva.ai marketing pages. Data: Marketing page views and clicks. Never loaded inside the signed-in app. Region: United States. Retention: Per Google's and Meta's terms. Only after you consent to analytics or marketing cookies.
Communications
- Resend: Every email Cresva sends: sign-in links, invitations, notifications, reports and the newsletter, through its API and its SMTP relay. Resend sends through Amazon SES.. Data: Recipient addresses and message content. Region: United States. Retention: Per Resend's terms.
- Google Workspace: The hello@cresva.ai mailbox: email you send us, including privacy and deletion requests, and email the team sends by hand. Data: The messages and their senders' addresses. Region: Per the vendor's data processing terms. Retention: Until deleted from the mailbox.
- Slack: Alerts and reports to the Slack workspace you connect. Data: Message content you asked to receive in Slack. Region: United States. Retention: Per your Slack workspace settings.
Payments
- Stripe: Subscription billing. Data: Billing name, email and payment details, and your account id. Region: United States. Retention: Per Stripe's terms.
Integrations
- Composio: Connections to third-party apps you choose to link. Data: The authorization you grant and the requests you make through it. Region: Per the vendor's data processing terms. Retention: Until you disconnect the app.
Research and search
- Google (Custom Search, PageSpeed and Chart APIs): Image search, page speed checks and QR code images. Data: Search queries, public page URLs, and the text of a QR code you ask for. Region: United States. Retention: Per Google's API terms.
- ScreenshotOne: Screenshots of public web pages for page analysis. Data: The public URL to capture. Region: Per the vendor's data processing terms. Retention: Cached by the vendor for 24 hours.
- Tavily: Web search for research and competitor monitoring. Data: Search queries. Region: Per the vendor's data processing terms. Retention: Per Tavily's terms.
- Brave Search: Image search. Data: Search queries. Region: Per the vendor's data processing terms. Retention: Per Brave's API terms.
- SerpAPI: Search result checks for AI visibility. Data: Search queries. Region: Per the vendor's data processing terms. Retention: Per SerpAPI's terms.
Security
- Cloudflare Turnstile: Bot protection on public forms. Data: A challenge token and your IP address. Region: Global. Retention: Per Cloudflare's terms.
Cresva, Inc. • 131 Continental Dr, Suite 305, Newark, DE 19713, USA