UlazAI Media plugin privacy policy

UlazAI Media processes only the data needed to connect your account, prepare one private 1K image, create it, and return the result. The plugin does not read your full chat history, sell your data, use plugin data for advertising, or publish your media.

Who is responsible for the data

UlazAI operates the UlazAI Media plugin and is the controller for data received by mcp.ulazai.com and ulazai.com. OpenAI separately controls the ChatGPT conversation and decides, under your ChatGPT settings, how long the copy in your chat is kept. This policy covers the data handled by UlazAI after ChatGPT calls a plugin tool.

Data used by each plugin step

Data	Why it is used	Who receives it	UlazAI retention
OAuth issuer and subject, verified email address, email-verification status, granted scopes, internal account ID, connection time, and last-use time	Connect one ChatGPT identity to one UlazAI account, authenticate requests, show the sponsored allowance, and block abuse	UlazAI and Auth0 by Okta	While the connection or UlazAI account remains active. Deleted after a verified deletion request, subject to the backup period below.
Prompt, optional negative prompt, selected model, aspect ratio, 1K resolution, privacy setting, idempotency IDs, quote ID, and job status	Validate the request, prepare fixed private parameters, prevent duplicate jobs, run the image model, and report progress	UlazAI and Kie.ai. OpenAI receives the tool result in your ChatGPT conversation.	A prepared request expires after five minutes and is removed by the next daily purge. Data for a started generation is kept for 30 days and removed by the next daily purge.
Generated private image, filename, MIME type, file size, dimensions, and SHA-256 checksum	Store one private result, verify the file, and create a five-minute download link for the user who requested it	UlazAI, Cloudflare R2, Kie.ai, and OpenAI when the private link is returned in ChatGPT	30 days, then removed by the next daily purge. Each signed download link expires after five minutes.
Request time, IP address at the network edge, user agent, rate-limit events, and restricted error/security codes	Deliver the service, rate-limit misuse, investigate failures, and protect accounts and infrastructure	UlazAI, Cloudflare, and the hosting provider	Application and security logs use three rotating files capped at 10 MB each per service and are purged by a server timer at least every 29 days. The size limit can overwrite them sooner. OAuth access-token contents are validated but not stored; a one-way token hash is kept only when needed to revoke a token and only until that token expires.
For a blocked or review-required prompt: safety category, matched term, risk score, action, and a prompt excerpt of up to 1,000 characters. A review may add reviewer notes.	Block unsafe generation, investigate abuse, and let an UlazAI moderator decide a borderline request	UlazAI moderators. Cloudflare processes an alert email when human review is required.	30 days, then removed by the next daily purge.

The daily purge may run up to 24 hours after a stated UlazAI deadline. Deleted database records can remain in encrypted backups for no more than 35 additional days before the backup rotation removes them.

Current tool inputs and outputs

Account: takes no user-entered fields. It returns connection and email-verification status plus whether sponsored image generation is available and how many sponsored images remain that month.
Model search: receives media type, operation, and optional aspect ratio, resolution, duration, audio requirement, quality or speed priorities, and result limit. It returns matching model names, descriptions, operations, supported formats, audio and input limits, and fit reasons.
Prepare image: receives an idempotency ID, operation, model, prompt, optional negative prompt, aspect ratio, resolution, optional duration, input-asset IDs, audio choice, and seed. The current public workflow rejects input assets and fixes the output to one private, unwatermarked 1K image with no Prompt Directory sharing. It returns a five-minute quote ID, the normalized parameters, model, media type, sponsored status, and expiry time.
Create image: receives the quote ID and a second idempotency ID. It returns the generation ID, status, operation, model, sponsored status, creation time, and suggested polling delay.
Get generation: receives the generation ID. It returns status and timestamps; on completion it also returns the media ID, filename, type, size, checksum, dimensions, and a five-minute download link. On failure a tool returns a restricted error code, message, retryability, and—only when account linking is required—a short-lived account-link URL.

Generation provider retention

UlazAI sends the prompt, selected GPT Image 2 route, and generation parameters to Kie.ai to create the image. Kie.ai states that generated media is kept for 14 days and text or metadata logs for two months, then deleted. Kie.ai does not receive your UlazAI password, OAuth access token, email address, full ChatGPT conversation, or payment data from this plugin flow. Read Kie.ai's current retention notice.

Data the plugin does not request

The public plugin does not request payment-card data, passwords, API keys, precise location, contacts, health data, government identifiers, or the full ChatGPT conversation. It has no checkout, paid balance, advertising, public gallery, social publishing, or automatic Prompt Directory sharing. Do not put secrets or sensitive personal data in an image prompt.

Processors and recipients

OpenAI: sends the tool input used for the requested operation and displays the tool output in ChatGPT.
Auth0 by Okta: authenticates the OAuth connection and provides the verified identity claims listed above.
Kie.ai: processes the prompt and generation parameters for the selected image route.
Cloudflare: protects the endpoints, stores the completed file in a private R2 bucket, and sends moderation alert emails.
UlazAI's hosting provider: runs the application, database, worker, and encrypted backups.

UlazAI does not sell plugin data or share it with advertisers or data brokers.

Your controls

You choose whether to connect UlazAI Media and whether to ask ChatGPT to prepare or start an image generation.
You can disconnect UlazAI Media in ChatGPT to stop future plugin access. Disconnecting does not by itself erase generation records already stored by UlazAI.
You can ask for access, correction, deletion, restriction, or a portable copy by emailing [email protected] from the email linked to the account. UlazAI verifies the request and responds within 30 days.
You can request early deletion of a private output instead of waiting for the automatic 30-day purge.

Legal basis, transfers, and age

UlazAI processes generation data to provide the service you requested. Account-security, rate-limit, and abuse-prevention data is processed for UlazAI's legitimate interest in keeping the service safe. The plugin workflow does not use marketing consent or behavioral profiling.

Some processors operate outside the European Economic Area, so plugin data may be processed internationally under the safeguards available for those services. UlazAI Media is not intended for children under 13. Do not submit a child's personal data or an image of a child through the plugin.

Effective date: 16 August 2026

Privacy and deletion contact: [email protected]