Privacy
Policy.

1. Who we are

Evil ASF, formerly known as Infinet, is operated by Clarity Digital Development, a trade name of VVS Vault LLC ("we", "us", "our"). Our chat product is available at evilasf.com, and infinetai.org continues to reach the same service. Contact us at [email protected].

The rename on 12 August 2026 changed the product’s name and nothing else. The data controller, what we collect, how long we keep it and your rights over it are all unchanged, and no personal information was transferred to a new party.

2. Scope of this policy

This Privacy Policy covers personal information we collect when you use evilasf.com or infinetai.org, sign up for an account, subscribe to a paid plan, or communicate with us. It also covers EvilASF Partner Tools, the Instagram app we run for our affiliates, including what it handles about people who comment on an affiliate’s posts (see 3.12). It does not cover third-party services we link to, which have their own privacy policies.

3. What we collect

3.1 Account information

When you sign up, our sign-in provider (Clerk) collects:

• Your email address, and a password hash if you set a password. You can also sign in with a one-time code we email you.
• If you sign in with Google, with Apple or with another account offered on the sign-in page, an identifier for your account there, and the email address it gives us
• If you sign in with Apple and choose to hide your email address, Apple gives us an address of its own, ending in @privaterelay.appleid.com, that forwards to yours. That relay address is the only one we see, and we use it as your email address.
• Profile fields you choose to provide (name, profile picture)
• Session metadata (IP address, user agent, login timestamps) for security

3.2 Payment information

If you subscribe to a paid plan, our payment processor (Stripe) collects and processes your payment card details, billing address, and transaction history. We never see or store your full card number. We receive a Stripe customer ID, the subscription status, and the last four digits for reference.

If you subscribe in our iPhone or Android app, Apple (through the App Store) or Google (through Google Play) sells you the subscription and charges you; we never see your card or your store account details. When you start a subscription in our Android app, the app also gives Google a one-way code made from your account ID. Google keeps it with the purchase to link the purchase to your account and to spot misuse. We share your account ID with a third-party subscription management provider, which receives your purchase records from the store and tells us your plan, its dates and whether it renews, the store’s transaction references, the amount and currency of each payment, any refund, and the country the subscription is associated with. Our iPhone app also sends that provider your device’s model, operating system, language and store country, the app’s version, and an identifier Apple gives our app on your device; our Android app sends the same without a device identifier. Like any service an app connects to, the provider also sees your IP address, from which it works out the country you are in; it records that country and when the app last connected to it. We use these records to give you your plan, where you joined through a partner’s link to pay that partner’s commission, and, as 3.10 describes, to count which of our ads lead to payments.

3.3 Usage data

We record basic operational data:

• Token counts per message (to enforce plan limits)
• Request counts per day (rate limiting)
• Model used per request
• Timestamps of each interaction
• Chat and message identifiers (anonymous UUIDs)
• Once a day, the network (IP) address, browser or app, and country you used, and which part of the service, so that we can answer a disputed card payment
• How many seconds of voice input and read-aloud you used, never the recording or the words
• For each Deep Research run, how many searches it ran and pages it read, its tokens and how long it took

3.4 Chat content (standard mode)

When you use the service in standard mode, the content of your messages is sent to Venice.ai, our AI inference provider, so it can write the replies. Each message is also checked by an automated classifier, on models Venice.ai hosts: on our website, for the only two things we refuse, sexual content involving a minor and sexual content depicting a real, identifiable person; in our apps, also for what the apps do not allow, which is sexually explicit writing, identifying a real person from a photo, and methods of self-harm, and there the model’s replies are checked as they arrive as well. When a message leads us to search our document archive, our server logs may record a short excerpt of that search (up to 50 characters) and the search terms derived from it, so we can diagnose search quality; our hosting provider keeps these logs for a limited period. Chat history is stored on your device, in your browser or in our app, not on our servers, with one exception in our app: if the app has not confirmed, within 10 seconds of a reply’s end, that it has the whole reply (for example because it went to the background, lost its connection or was closed while the reply was arriving), we keep that reply as the app would have shown it, except for the model’s thinking and the search terms and file names its tools were given, so that the app, if it is still running when you come back, can show it. We delete it as soon as the app confirms it has it; otherwise it is kept for 15 minutes and then deleted by a job that runs every hour, normally within the hour after that, about 75 minutes after the reply ends. We never do this on our website or for a Private Mode conversation. We do not train AI models on your content. We do not sell, rent, or share your content with advertisers or data brokers.

To give a chat its title, its first message is sent to Venice.ai, which suggests one. If a message carries a link, our server fetches the page through Jina Reader, which receives the page’s address, and passes what it read to Venice.ai with your message. If you turn on web search, Venice.ai searches the web for the reply.

Voice. If you use voice input, your recording is sent to Venice.ai to be turned into text, and if you use read-aloud, the reply is sent to Venice.ai to be spoken. We keep neither the recording nor the words, only how many seconds you used, which count against your plan.

Deep Research. A Deep Research question is sent to Venice.ai, which plans the research and writes the report. Its searches go to Brave and to the public forums named in section 5, and the pages it reads are fetched through Jina Reader. Unlike a chat, a report is stored on our servers, with your question and the sources it used, so that you can open it again. It stays until you delete it or your account.

Photos you attach in our mobile apps are sent on without their location or any other metadata embedded in the file: our server removes it before a photo is forwarded to an AI provider.

3.5 Chat content (Private Mode)

When you enable Private Mode (available on VIP and above), your conversation is answered only by models that run inside sealed, hardware-attested computing (a trusted execution environment); if none is available, no model answers, and you are told so. Unless you report a reply from it (see 3.14), nothing from a private conversation is kept on our servers: we do not write its content to our logs, nothing from it is saved to memory or shared, a conversation that has been private stays private, and your chat history stays on your device, in your browser or in our app. Your saved memories are not used, and a private chat is titled on your device from its first words. Web search, link previews, voice input and read-aloud are switched off in Private Mode, because each would send your words to a service outside sealed hardware, and so is Deep Research, whose reports are stored. The model that answers in Private Mode cannot see images, so a photo attached to the message you send is refused before any model sees it, and a photo earlier in the conversation is replaced by a note saying one was there.

Private Mode is not end-to-end encrypted. Our servers receive each message in readable form, in memory, in order to route it, to search our document archive, and to screen it: on our website for the two things we refuse, and in our apps for what the apps do not allow, along with the model's replies as they arrive. That screening also runs on sealed-hardware models. Our inference provider also receives the message in readable form and relays it to the sealed hardware. We retain only metadata (timestamps, token counts, model used) for billing and operational purposes.

3.6 Community chat

Community rooms and direct messages are stored on our servers. This is different from your conversations with the AI, described above, which stay in your browser. A chat room cannot function without server-side storage, so if you use community chat, assume the messages are retained.

Community messages are not end-to-end encrypted, and Private Mode does not apply to them. We can access message content in rooms and direct messages, including private rooms, and we do so where it is necessary to investigate a report, respond to automated safety screening, enforce the Terms, or comply with a legal obligation. Messages are screened automatically for a narrow set of serious harms — sexual content involving minors, trafficking, credible threats of violence, publishing a private person’s identifying details, and indications of self-harm. We do not screen for opinions, language, or subject matter.

The handle you choose is visible to other members. Deleted and removed messages are hidden from the room but retained in our records, because a deleted message is often the evidence for the report about it. If you do not want a message stored on our servers, do not post it in community chat.

Cipher pages. When you record a reading on one of our Cipher pages, we store your transcription of the page, your method and the plaintext it gives, under the community handle you have at the time, and compare it with other members’ readings to see whether they agree. If you ask for a model’s reading of a page, the page’s image is sent to Venice.ai, and its reading is shown to you alone and is not stored.

3.7 Shared chat links

If you use the Share button on a conversation, a copy of that conversation is stored on our servers so the link can be opened by whoever you send it to. Apart from a report you choose to send (see 3.14), this is the only case in which your chat with the AI is stored by us, and it happens only when you choose it. The dialog says so before the link is created.

What is stored is a snapshot taken at the moment you shared — messages you send afterwards do not appear on a link you already shared. System instructions, saved memories and any images in the conversation are not included. The link is a random, unguessable address; the page is not indexed by search engines, but anyone you give the link to can read it and can pass it on.

Shared links expire automatically after 30 days and are deleted from our servers. You can revoke one sooner at any time in Settings, which deletes it rather than merely hiding it. Deleting your account deletes all of them. If a shared chat is reported, the report keeps its own copy for the time 3.14 sets out. Private Mode conversations cannot be shared: sharing stores a copy on our servers, and nothing from a private conversation is kept there unless you report a reply from it.

Anyone who opens a shared link can report it to us (see 3.14). While we review a report that a chat shows sexual content involving a minor or depicting a real person, the link may stop opening, and your list of shared links in Settings shows it as under review.

3.8 Memory (opt-in)

Memory is off by default. Your conversations are stored in your own browser and are not sent to us for storage. If you switch memory on in Settings, we store short factual statements — either ones you type yourself, or ones the assistant saves when you tell it something it judges durable, such as a standing preference or a subject you are researching. We do not store the conversations those facts came from.

Everything saved is listed back to you in Settings, in full and in plain text. You can delete any entry individually or erase all of them at once. Switching memory off stops us reading or writing memories immediately but does not delete what is already stored — use “forget everything” for that. Deleting your account removes all of it.

3.9 Diagnostic data

We collect basic server logs (IP address, request path, status code, user agent, timestamp) for security, abuse prevention, and debugging. These logs are retained for up to 90 days.

When something fails while serving you, our error log records the failure, which may include your account’s identifier, so that we can fix it. When you sign up, we note the network address you signed up from and the domain of your email address, to stop many accounts being made from one place.

Updates to our apps. Our iPhone and Android apps receive their updates from Expo, the service they are built with. Each time one of our apps starts, it asks Expo’s update service (u.expo.dev) whether an update is waiting, and downloads it from Expo if one is. Each check and each download sends the app’s version, whether it runs on iOS or Android, its update channel, which update it is running and which it was built with, and a random identifier the app makes for its installation. On an iPhone, the system also adds what it adds to any app’s connections: the app’s name and build number, the iOS version and your language settings. Expo’s service can also ask the app to send a value of its choosing back with later checks. The identifier is kept with the app’s settings on your phone until you delete the app. Like the app’s other settings, it is included in your iPhone’s backups and comes back on a phone restored from one, and on Android it moves with a direct phone-to-phone transfer. Expo uses it to count the installations that check for and receive each update. If the app cannot load an update as it starts, or, on an iPhone, crashes as it starts or in its first seconds, the next check also sends the error message that stopped it (cut to about 1,000 characters), and until newer updates replace them, each check also sends which recent updates failed to start, so that Expo can report how many starts fail. A check carries no name, email address or account identifier, and we do not link the identifier to your account, so deleting your account cannot reach it; deleting the app removes it from your phone. Like any service an app connects to, Expo sees your IP address.

3.10 Cookies and local storage

Clerk, our sign-in provider, sets cookies that keep you signed in. When you pay, you are sent to Stripe’s own checkout page, and Stripe sets its cookies there, on its own site. Stripe’s payment script also loads on our pricing page and inside the app, so that a checkout can start, and it sets its own fraud-prevention cookies on our site as well. Your browser’s storage keeps your chat history on your device, plus a few settings, such as whether you have already seen the tour. The cookies we set ourselves are described next: the ones that measure where visitors come from, and one for pages opened from our app.

Where visitors come from (first party). If you arrive from a creator’s referral link, from one of our ads, from a link we send you in an Instagram message, or from any other link to us that carries campaign tags, we store which link, ad or post brought you, in cookies we set ourselves: inf_ref, inf_ref_at and inf_ref_hit for referrals, and ad_touch, ad_touch_ft and ad_hit for everything else. The referral cookies last 60 days and the others 90 days, except inf_ref_hit and ad_hit, which last ten minutes.

A link we send you in an Instagram message, for example after you comment on one of our posts, is evilasf.com/t/ followed by a number. That number identifies the Instagram post the message was sent about, not you. Every tap on the link is recorded, including a repeat tap, and so is the app loading the page in advance. When the message comes from the account of a creator in our affiliate program, the link also ends with that creator’s referral code, which credits the creator and sets the same referral cookies as their other referral links. A creator with a short link of their own on our site may send evilasf.com/ followed by their name and a short code instead; that code is the same post number written in fewer characters, and the link works in exactly the same way.

When you arrive through one of these links, we store a visit record: which link, ad or post it was, the page you landed on, the time, and a scrambled (hashed) form of your IP address. For every link except a creator’s referral link, the record also notes the country your connection comes from and whether the link opened inside an app’s built-in browser. That record does not name any account.

We use the referral cookies to pay creators what they are owed, and the others to count which ads and posts lead to sign-ups and payments. If you arrived from an ad, a message link or another tagged link and then create an account in the same browser, we note on your account which one brought you and when you arrived (the first and the latest, if there was more than one). That note is deleted with your account. The cookies themselves stay in your browser until they expire or you clear them.

Pages opened from our app (first party). When our mobile app opens one of our pages, it adds a code to the page’s address that our server signed for it and that works for two minutes; such a code, and nothing else, makes us set a cookie named app_shell, holding only the time it was set. While it holds (at most 30 minutes), our front page, our plans, sign-in, sign-up, billing and the website’s chat show a page asking you to go back to the app, and our terms, this privacy policy, the data deletion page and shared chats leave out their own links to them. It is strictly necessary for that, and we use it for nothing else. It is a session cookie, so your browser deletes it when it closes, and we ignore it after 30 minutes. Opening any of our pages with ?app=0 in its address removes it.

What we do not do. We do not use Google Analytics, the Meta Pixel, the TikTok Pixel, or any other advertising tracker. The only outside analytics service we use is PostHog, on the public pages listed in 3.11 below and only in the way described there. Apart from it, Expo counts our apps’ installations and failed starts from their update checks, as 3.9 describes. The chat app does not start PostHog, and it loads no advertising code. The ad and message-link records described above stay on our own servers: we do not send your sign-ups, your payments, your chats or any identifier back to TikTok, Meta or any other advertising network, and we do not track you across other websites.

3.11 Website analytics (PostHog)

On some of our public pages we use PostHog, a website analytics service, to see how those pages are read: which links and posts bring people to them, how long people stay, and how far down they scroll. We use PostHog’s cloud in the European Union, at eu.i.posthog.com.

Where it runs. Only on these pages of evilasf.com: the home page, pricing, the coverage page, our philosophy page, the uncensored AI page, the surveillance pages, the Flock page, the blog, this privacy policy, the terms of service, the API terms, and the main affiliate program page. No other page loads it, and it records nothing from any other page. That includes the chat app, community, your account and billing pages, sign-in and sign-up, shared conversations, the document archive and the cipher pages, the affiliate dashboard, handbook and terms, and the form for reporting intimate images. A link from one of the pages above to any of these opens it as a fresh page, so PostHog does not come along.

What it records. Each time you open one of the pages above, and when you leave it, PostHog receives:

• The page’s address, cut down to the page itself and its campaign tags (the utm_ tags that say which link brought you). Anything else in the address is removed before it is sent.
• The website that sent you, by its domain name only, never the page you were on there. When you came from another of our own pages, this is left blank.
• How long you stayed on each of the pages above, and how far down you scrolled.
• Your browser and its version, your operating system and its version, whether you are on a phone, tablet or computer, and the size of your screen and browser window.
• The technical details PostHog needs to file each event: the time, random identifiers for the visit and the page, our PostHog project key, the name and version of PostHog’s code, and a flag telling it not to build a profile.

That is everything our code sends. If you came from a link we sent you in an Instagram message (see 3.10), the campaign tags include the number of the Instagram post the message was about, so PostHog sees that number too. It identifies the post, not you.

What it does not do. PostHog sets no cookies and saves nothing in your browser’s storage. Its identifier for you is random and lasts only as long as the page stays open. Moving from one of the pages above to another usually keeps the same one, but reloading the page, or coming back later, starts a new one. So PostHog does not connect one visit to the next. It is never told who you are, it builds no profile of you, and it does not link any visit to your account. It does not make a screen recording or replay of your visit, and it does not capture your clicks or your typing. It makes no heatmaps, shows no surveys, and fetches no settings that could switch any of that on. Like any website your browser connects to, PostHog sees your IP address when an event arrives. It is set to discard it rather than keep it, and to work out no location from it.

Do Not Track and Global Privacy Control. If your browser sends either signal, PostHog is not loaded at all, and nothing is sent to it.

3.12 Instagram replies for affiliates (EvilASF Partner Tools)

EvilASF Partner Tools is an Instagram app we run for our affiliates, at their request. An affiliate connects their Instagram account and chooses a keyword and a message. When someone comments that keyword on one of the affiliate’s posts, the app sends that person the affiliate’s message as an Instagram direct message. It runs as a separate service from evilasf.com.

If you comment on an affiliate’s post, the app stores:

• Your Instagram username
• Your Instagram-scoped ID: the number Instagram gives the app for your account
• The text of your comment
• The message the app sent you, and Instagram’s receipt for it
• The times of your comment and of the reply

If you are an affiliate, it stores:

• Your Instagram username and account IDs
• Your email address
• An access token that lets the app read the comments on your posts and send messages from your Instagram account, stored encrypted
• The keyword that triggers a reply, and the text of the message it sends
• Which version of the Partner Tools agreement you accepted

What it is used for. Information about the people who comment is used only to send that affiliate’s reply. It is kept separately for each affiliate, and it is never used for our own marketing.

What reaches evilasf.com. Only totals reach evilasf.com: for each post, how many messages were sent and how many failed each day, with the post’s ID and web address. Never a commenter’s ID or username, their comment, or a message ID. If the message you were sent includes a link to evilasf.com and you tap it, that visit is recorded as 3.10 describes.

How long it is kept.

• A commenter’s username, Instagram-scoped ID and comment: 30 days after the reply reaches a final status (sent, or failed for good), then deleted. What stays is a scrambled (hashed) key, the reply’s status and times, and the ID of the post, none of which includes the username or the comment.
• Message and recipient IDs from Instagram’s receipt: 30 days.
• The requests the app sends to Instagram to deliver a message, and Instagram’s responses: 30 days.
• Logs of the app checking for new comments: 30 days.
• Alerts to our staff about problems in the app: 90 days after the problem is resolved. Alerts carry no information about commenters.
• Links we send an affiliate to connect their account, and the sign-in state used while connecting: 7 days after they are used or expire.
• An affiliate’s access token: until they disconnect or ask us to delete their data, then deleted at once.
• When an affiliate disconnects: what the app holds about the people who commented on their posts is deleted 24 hours later. The daily totals stay.
• An affiliate’s other details (username, account IDs, email address, keyword, message and agreement version): until they ask us to delete them.
• Our staff’s audit log of what they did in the app: 1 year.
• Deletion requests: 1 year, as a reference code and a status only.

Disconnecting and deletion. An affiliate disconnects by removing the app in Instagram’s settings, or by asking us. Instagram tells the app when it is removed: the app deletes the affiliate’s access token at once, and what it holds about the people who commented on their posts 24 hours later. Instagram may also pass on a request to delete the affiliate’s data. Then the app also deletes their Instagram username and account IDs at once, along with what it holds about those commenters. Anyone, whether an affiliate or someone who got a message from the app, can ask us to delete what it holds about them by emailing [email protected] with their Instagram @username, and we will do so within 30 days. Our data deletion page gives the steps and says what we keep afterwards: the daily totals, and for affiliates the commission and payout records we must keep for tax and accounting (see section 6).

3.13 Consent in our apps

Before you send your first message in our apps, we ask you to confirm that you are 18 or older, to agree to how your messages are processed by the companies named in section 5, and to accept our Terms of Service, having read this policy. We record each answer: which question it was, the version of the words you saw and a fingerprint (hash) of them, when you answered, and which app and version you used, so that we can show which words you agreed to. When the words change in a way that matters, the app asks you again. These records are kept while your account exists and are deleted with it.

3.14 Reports

Reporting a reply in our apps. When you report a reply, the app sends us the reply, the message it answered (a photo in that message is counted, never sent), the reason you chose and any words you add, and, only if you tick the box, up to six earlier messages from the conversation. We store the report with your account, which app and version you used, which model replied and whether Private Mode was on. A report of a reply from a Private Mode conversation is stored the same way, and the report sheet says so before you send it.

Reporting a shared chat. Anyone can report a shared chat from its page, signed in or not. We store a copy of the shared conversation as it was when you reported it, the messages you point to, the reason you chose, your words and, if you give one, an email address for our reply. If you are signed in, we store the report with your account; if not, we store your network (IP) address and browser, to prevent abuse. Either way we keep a scrambled (hashed) form of your network address, which we use to count how many different people have open reports of the same chat as sexual content involving a minor or depicting a real person, and to file a further report of the same chat, for the same reason, from the same network, while the first is still open, without alerting our staff again. To stop a flood of reports, we also count the reports sent from each network in each hour, under another scrambled form of its address, and delete those counts after a day.

What happens next. Our staff review each report. We email them when one arrives, without what was reported, but not again for a further report of a shared chat for the same reason from the same network while the earlier one is open. A report that a shared chat shows sexual content involving a minor or depicting a real person can stop the link opening while we review it, when the reporter is signed in with a verified email address, when more than one person has reported it, or when our automated check agrees. A report of a shared chat as “Sexual content about a real person” also goes to the process described on our reporting page: it is filed with our intimate-image reports with your words, any email address you typed, your account identifier if you are signed in, and the link to the shared chat, and the team that handles them is emailed the same, except for a further report from the same network, as above. The report keeps your browser only if you were not signed in. A report of a reply in our apps is not filed there. We record what we decided.

How long it is kept. What was reported, your words, your account identifier, your email address and your network details are deleted 90 days after the report is resolved, or a year after for a report of sexual content involving a minor or of trafficking. A legal hold keeps all of them until the hold ends; and, held or not, what was reported also stays while a later report of the same chat still relies on that copy. The record of the report and of what we decided stays. A report of a shared chat as “Sexual content about a real person” is kept with our intimate-image reports as that report’s record, with the words, email address, account identifier and link described above and, if you were not signed in, your browser; it is not emptied on this 90-day schedule.

If you delete your account. A report of sexual content involving a minor or of trafficking, and any report under a legal hold, is kept exactly as it was made until the time above. Any other report you made of a reply is deleted if our staff had already dismissed it or found it a duplicate when you delete your account. One still open then, or one they acted on, keeps your account identifier until the time above, even if they dismiss it later. Any other report you made of a shared chat loses your account identifier, email address, network address and browser at once; the scrambled form of your network address described above stays with it, for the same uses, until the time above. Any other report about a chat you shared stops naming your account if our staff had already dismissed it or found it a duplicate when you delete your account. One still open then, or one they acted on, keeps your account identifier until the time above, even if they dismiss it later. If you reported a shared chat as “Sexual content about a real person”, its filing with our intimate-image reports keeps your words, the link and any email address you typed, which our reply goes to, and loses your account identifier, whatever happens to the report itself.

4. How we use your information

• Operate the service: Route your messages to AI providers, return responses, enforce plan limits.
• Safety: Screen messages and replies as 3.4 and 3.5 describe, and review reports and act on them (see 3.14).
• EvilASF Partner Tools: Send the Instagram reply an affiliate has set up to the people who comment on their posts. What the app holds about those people is used for nothing else (see 3.12).
• Billing: Process subscriptions, handle upgrades/downgrades, manage refunds when applicable.
• Security and abuse prevention: Detect and block fraudulent accounts, abusive patterns, or violations of our Acceptable Use Policy.
• Communication: Respond to support requests, send transactional emails (billing receipts, subscription changes, service announcements).
• Legal compliance: Respond to lawful requests from courts or regulators where required by law.

We do not use your content to train AI models. We do not sell your personal information. We do not share your content with advertisers.

5. Third-party processors

We rely on the following third parties to run Evil ASF, and each receives what its line describes. Each has their own privacy practices and certifications:

• Venice.ai — answers your messages, titles chats, writes Deep Research reports, turns voice into text and replies into speech, reads Cipher pages you ask about, and runs our safety and community screening, on models it hosts; in Private Mode, only its sealed-hardware models answer (see 3.4 and 3.5)
• Brave — web searches for Deep Research
• Jina Reader — fetches the pages a message links to, and the pages Deep Research reads; it receives their addresses
• Forum APIs (Hacker News via Algolia, Stack Exchange, Discourse at discuss.privacyguides.net, 4chan and 4plebs) — searched by Deep Research, with the search terms it chooses
• Clerk — sign-in and session management (see 3.1)
• Stripe — payment processing, billing and subscription management (see 3.2)
• Our subscription management provider — keeps the records of subscriptions bought in our iPhone and Android apps through the App Store or Google Play, and tells us which plan each account has (see 3.2)
• Expo — delivers the updates to our iPhone and Android apps, which ask it for one each time they start (see 3.9)
• Our hosting provider — runs our website and the servers behind it, for evilasf.com
• Our database provider — hosts the database everything we store for evilasf.com is kept in
• Cloudflare — the network in front of evilasf.com, which carries every request to us and protects the site from attacks
• Resend — sending our emails, including those from EvilASF Partner Tools
• PostHog — website analytics on our public marketing pages, hosted in the European Union (see 3.11)
• An alerting service — tells our staff when EvilASF Partner Tools needs attention; its alerts carry no information about the people who comment

When you use features like web search, voice or Deep Research, what the feature needs is sent to the companies named above for it. We do not control their data practices beyond our agreements with them; the forum APIs are public services that Deep Research searches under their own terms.

6. Data retention

• Account data: Retained for as long as your account is active. You can delete your account yourself at any time under Settings → Account; this removes your account data immediately rather than on a request queue, and cancels any subscription you pay for on our website at the same time. It does not cancel a subscription bought through the App Store or Google Play; cancel that in your App Store or Google Play subscription settings. You can also delete it without signing in: our account deletion page emails a link that deletes it, only to an address your account has confirmed and only three a day to one address, and lists every kind of data deleting your account deletes and every kind it keeps, and for how long. While you ask, we count requests from your network and to your address, by scrambled forms of them, for a day, and keep a record of the link we sent, without the link itself, until a day after it is used or lapses. A job that runs every hour deletes them once that day is over.
• Account records: A record of the emails we sent you, the usage alerts we sent you, the history of your plan changes and the trial offers you had are kept while your account exists and deleted with it. A coverage request you sent is kept as an area request only, with your account identifier and contact details removed.
• Payment records: Retained as required by applicable tax and accounting law (typically 7 years).
• Usage records: The token count, model and time of each message are kept while your account exists, because your plan’s limits and our billing are counted from them, and are deleted with your account. The request times we keep for rate limits are deleted after 30 days, and the daily record of your network address, browser or app, and country after 90 days.
• Chat content (standard mode): Stored on your device, in your browser or in our app. In our app, a reply the app has not confirmed it has in full within 10 seconds of its end (for example because the app went to the background, lost its connection or was closed) is kept on our servers, without the model’s thinking or the search terms and file names its tools were given, so that the app can show it if it is still running when you come back, until the app confirms it has it, and otherwise for 15 minutes, then deleted by a job that runs every hour, normally within the hour after that, about 75 minutes after the reply ends; never on our website or in Private Mode. Deleting your account deletes it too. You can delete a chat at any time, on our website or in our app, and clearing your browser’s storage removes all of it from that browser.
• Chat content (Private Mode): Not retained server-side, unless you report a reply from it (see 3.14). Kept only on your device, in your browser or in our app. Cannot be shared as a link.
• Shared chat links: A snapshot of the conversation you chose to share, stored on our servers and deleted automatically 30 days after you created the link. You can revoke a link sooner from Settings, which deletes the snapshot immediately. Deleting your account deletes all of them.
• Deep Research reports: Kept, with your question and the sources each used, until you delete the report or your account.
• Voice: Only the seconds you used, kept while your account exists and deleted with it.
• Consent records: Kept while your account exists, and deleted with it (see 3.13).
• Reports: What was reported, the reporter’s words, account identifier, email address and network details are deleted 90 days after the report is resolved, or a year after for a report of sexual content involving a minor or of trafficking; a legal hold keeps all of them longer, and, held or not, what was reported also stays while a later report of the same chat still relies on that copy; the record of the report and of our decision stays (see 3.14). A report of a shared chat as “Sexual content about a real person” is filed with our intimate-image reports and kept there as that report’s record, not emptied on this schedule. The counts that stop a flood of reports are deleted after a day. Deleting your account deletes some reports at once and takes your account identifier off others, as 3.14 sets out. An intimate-image report you made is kept as the legal record of the report, with your account identifier removed: where the content is, what you told us, when it reached us, the network address and browser it came from where we recorded them, and any email address you gave, which our reply goes to, all stay with it.
• Community messages: Retained for 12 months, then deleted. Messages that are subject to a report or flag our staff have not resolved, or that we are required to preserve for a legal obligation, are kept until that is concluded. Deleting your account removes your membership of any room, and your handle stays reserved, so no one else can take it; it does not delete messages you have already sent to other people, which remain part of those conversations under the handle you used, without your account identifier. A message under a report we passed to the authorities, or one our automated check flagged as sexual content involving a minor or as trafficking in the last year, keeps your account identifier until it is deleted, and so does a report you made that we passed to the authorities. Private rooms you own stay open to their other members.
• Cipher readings: Kept as the record of each page, under the community handle you had when you recorded them. Deleting your account removes your account identifier from them and from your votes; the readings and the counts stay.
• Error records: Kept so that we can fix the failures they record. Deleting your account removes your account’s identifier from them.
• Sign-up screening: Deleted with your account, except that if your account is less than a day old, the network address it signed up from is kept for up to 8 days, without your account or email domain, to stop repeated sign-ups from one place.
• Legal hold: If, in the year before you delete your account, a chat you shared was reported to us as sexual content involving a minor or as trafficking, or our automated check flagged something you posted, or tried to post, in the community as either, and that report or flag was not dismissed, the email addresses verified on your account and the network address you signed up from are kept for a year from that report or flag, for the authorities, and then deleted. If our staff dismiss every such report and flag, they are deleted then.
• Affiliate partners: If you were an affiliate partner and delete your account, your referral code and link stop working, commissions you had already earned are still paid, and your name and email address stay with the commission records for 7 years after the deletion, for tax and accounting, and are then removed.
• EvilASF Partner Tools: Kept for the periods listed in 3.12; a commenter’s username, ID and comment, for example, are deleted 30 days after the reply reaches a final status. To have it deleted sooner, see our data deletion page.

7. Your rights

Depending on your jurisdiction, you may have the right to:

• Access — request a copy of the personal data we hold about you.
• Correct — ask us to fix inaccurate data.
• Delete — delete your account and its data yourself from Settings → Account, ask for a link on our account deletion page, or ask us to. Either way we act subject to legal retention requirements (e.g. tax records held by our payment processor). For EvilASF Partner Tools, including if you received a message from it, email [email protected] with your Instagram @username; our data deletion page says what happens next.
• Object — object to our processing of your data.
• Portability — receive your data in a portable format.
• Withdraw consent — where we rely on consent, you can withdraw it anytime.

To exercise these rights, email [email protected]. We will respond within 30 days.

8. Children

Evil ASF is not intended for users under 18. We do not knowingly collect personal data from children. If you believe a child has created an account, please contact us and we will delete the account.

9. International data transfers

Our services are hosted in the United States. If you access Evil ASF from outside the US, your data will be transferred to and processed in the US and potentially in other jurisdictions where our service providers operate. By using the service, you consent to this transfer.

10. Security

We use industry-standard safeguards including HTTPS/TLS for all data in transit, encrypted storage at rest for databases, and access controls on internal systems. Private Mode runs the model inside sealed, hardware-attested computing and keeps its content out of our logs and storage, unless you report a reply from it; it is not end-to-end encrypted (see section 3.5). However, no system is perfectly secure. If we become aware of a breach affecting your data, we will notify you within the time period required by applicable law.

To report a security vulnerability in evilasf.com or in EvilASF Partner Tools, email [email protected] with enough detail for us to reproduce it.

11. California residents (CCPA/CPRA)

If you are a California resident, you have additional rights under the California Consumer Privacy Act including the right to know what personal information is collected, the right to delete personal information, the right to opt out of sale or sharing of personal information (we do not sell or share), and the right to non-discrimination for exercising these rights. To exercise any of these, contact us at the email above.

12. European residents (GDPR)

If you are in the European Economic Area, United Kingdom, or Switzerland, our legal bases for processing are: performance of contract (providing the service you signed up for), legitimate interests (security, abuse prevention), legal obligation (tax records), and consent (optional features). You have the right to lodge a complaint with your local data protection authority.

13. Changes to this policy

We may update this policy occasionally. Material changes will be announced on this page with an updated effective date, and for significant changes we will email registered users. Your continued use after a change constitutes acceptance.

14. Contact

Questions, requests, or concerns: [email protected]