Version: v0.7 • Effective: 1 September 2026
Who we are: Taca is an AI-powered assistant and CRM operated by Tacatech Ltd, a company registered in England & Wales under company number 17355560, with its registered office at 71–75 Shelton Street, Covent Garden, London, WC2H 9JQ, United Kingdom.
Contact: privacy@tacatech.net
Data protection registration: Tacatech Ltd is registered with the UK Information Commissioner's Office (ICO) under registration reference ZC205697.
Taca is an AI-powered assistant and CRM for Fanvue creators. This Privacy Policy explains what personal data we process, why we process it, and how we protect it.
Where Taca processes fan data on behalf of creators, the creator is generally the data controller and Tacatech Ltd acts as the data processor where applicable under the relevant agreement. Taca is an independent platform and is not affiliated with, endorsed by, or operated by Fanvue.
About you (the creator):
Taca account tied to it, your plan and billing status.
encrypted at rest (AES-256-GCM). We never see your Fanvue password.
before auto-send, each auto-reply's own draft text is held briefly (typically seconds
to a few minutes) so you can see it, edit it, or cancel it before it sends - the one
other place a draft's text is briefly stored, alongside the audit log above. It's
deleted the instant it sends, is cancelled, or fails, with a strict 24-hour backstop
regardless. Off by default; nothing is held unless you turn it on.
never see, receive or store your card details. We only read from Fanvue *which plan
you are on*, so we know which features to unlock.
Facial and biometric images - the Studio is switched off:
images. It ships disabled - a single kill switch (STUDIO_ENABLED) that is OFF in
production. While it is disabled, Taca does not ask for, collect, store, generate or
train on any photograph of your face or body, and we hold **no special-category
(Article 9) biometric data** about you.
gated on this policy and the DPA being updated first, a DPIA being completed, and
explicit in-app consent (that you own the rights and every subject is 18+). Until then,
any reference below to "facial processing" describes what the feature would do once
enabled - not something happening today.
What we read for you (via Fanvue, with your approved permissions read:self,
read:chat, write:chat, read:creator, read:fan, read:insights, read:media,
plus openid and offline_access for signing you in and keeping your connection
active):
your fans; you are their controller and we process them for you (see the DPA).
We do not keep our own copy of your full conversation history for display.
keep a log of what Taca did: each draft prepared, message approved/sent, and message
blocked or flagged. An entry can include the fan's display name and the **draft or
message text involved. This content is redacted after 90 days** (the action record
and its date are kept for your activity totals), and the whole log is deleted when you
delete your account. You can export it (Settings → Export my data).
perform better over time, Taca keeps some data about how your account performs,
separate from your live inbox:
one that failed or wasn't sent), recording which settings were behind it (persona
style, selling style, which AI provider) and a **de-identified reference to the
fan (a one-way hash, not their real Fanvue id). Never the message text.**
unsubscribed, linked back to the reply that came before it. Again, never message text.
conversations to teach Taca your voice, we run one analysis pass and keep only short
structured findings about how you write (tone, pacing, selling style) - never the
conversation text itself, which is deleted immediately once analysed (or within 24
hours if analysis fails).
gently inform future drafts, the same way the "learned voice" from your sent messages
already does - your own explicit persona settings always take priority over it.
There is no dedicated in-app screen to view or turn this off per account yet; you can
ask us to disable it for your account, or remove it entirely by deleting your account.
have performed against each other (for example, one persona style outperforming
another). Never applied automatically - you review and choose to apply or dismiss
each one.
automatic-rollout capability exists in Taca's backend, but **has no interface in the
product yet** and is switched off by default. No creator account is currently running
an experiment or automatic optimisation through Taca.
None of this is shown to, or shared with, your fans, and all of it is deleted when you
delete your account.
"Help improve Taca" (your own genuinely human-written material only). Separately
from the Learning Engine above, Taca keeps short, identifier-scrubbed windows of
conversation turns from your account (personal identifiers such as emails/phone
numbers/handles are stripped before storage; never structured against a specific fan
by name). This is governed by a dedicated, separately auditable setting called
"Help improve Taca." Because your Fanvue connection can give Taca access to
conversations that took place before you ever connected Taca, **this setting is shown
to you before you connect your Fanvue account**, not only afterward - on by default,
with an immediately available off switch, and the same setting stays available any
time after in Settings → Help improve Taca:
to be reviewed, alongside other opted-in creators' own material, to refine the fixed
behavioural rules and reference material Taca's replies draw on (never used to
fine-tune an underlying AI model). This can include **eligible past conversations
reachable through your Fanvue connection**, not only messages sent after you turn
this setting on - we do not begin using anything for this purpose, historical or
new, until your actual choice is known. **A reply Taca itself drafted and sent from
your account is never used as this kind of evidence, even while this setting is on**
anything we cannot confirm either way is excluded too, not assumed in your favour.
Your fans' own messages are never used for this cross-creator purpose, regardless
of your setting - fan-authored turns have their own, completely separate eligibility
gate, off by default, that your setting cannot switch on.
under it is deleted. Turning it off never changes how Taca works for you.
collected and left unused** - once your account is eligible (this setting on, and
your account genuinely holding active shareable status), your own confirmed-human
material is actually reviewed alongside other opted-in creators' material. See §9A
for the data-protection assessment status this was enabled ahead of.
product for the benefit of the wider creator base, balanced against the design
choices described in §9A. This setting is a **product control, not itself how we
establish a GDPR legal basis** - we do not describe switching it on as "consent."
its history (when it changed, and under what version of this notice) is recorded for
your own account.
Our secondary database (Supabase) - metadata and account details:
We run a secondary database (Supabase; see the sub-processor table) that holds a
rebuildable cache of metadata plus a mirror of your account details:
handle, Fanvue-account id and active/inactive state. **Your encrypted Fanvue tokens are
NOT copied here** - they stay only in our main database.
in your inbox), a top-spender flag, and last-seen time.
its Fanvue id, direction (incoming/outgoing), category (normal / mass message / Fanvue
automated), timestamp, and whether it had an attachment (and how many).
media links, avatars, or pay-per-view content or prices.
deleting it loses nothing, and each part can be switched off without affecting the app.
immediately, and the same action then deletes it from this secondary database automatically
(one delete of your account row, cascading to the rest). If that second step fails (for
example the database is unreachable), the deletion of your main record still stands and the
failure is recorded so we can complete the removal; you can also ask us at any time (see
Your rights).
To run the app, draft your replies, screen every message for safety/compliance, run any
automation you switch on, support you, and meet legal duties. Lawful bases: contract
(running the service), legitimate interests (security, prevention of abuse,
improvement), and legal obligation (safety screening, record-keeping).
We share data only with the providers needed to run Taca. This table is accurate to
the current system - update it whenever a provider is added or swapped.
| Service | Used for | Data it may see | Region |
|---|---|---|---|
| Fanvue | Your account, inbox, fans, media, earnings. Also takes all payments - see below. | Everything you connect via OAuth | Per Fanvue's policy |
| OpenRouter | Default provider for writing your inbox fan-reply drafts. Configured to route to an uncensored open-weight model (currently Hermes). | Fan message text, persona settings, conversation history (text only - no images) | US |
| Venice AI | Default provider for check-in re-engagement drafts, persona voice drafts, persona feedback, and fan-memory extraction; fallback for inbox fan-reply drafts specifically if OpenRouter is unavailable; also captioning a vault image you ask us to caption (2026-09-01: corrected — Venice was previously described here as "fallback only," which was accurate for the inbox reply draft but not for these other features, where it is the default) | Fan message text, your persona settings, conversation history, and the vault image being captioned (may be explicit) | US |
| Supabase | Secondary database: account details + rebuildable metadata cache (see §1) | Your email + account details; fan/conversation/message metadata only - no message text, media or tokens | See §8 |
| App host - Fly.io | Running the app + main database | Everything stored by the app | London (LHR) |
We do not take payments. Taca is a Fanvue App: your subscription is bought and
managed inside Fanvue, and we never see, receive or store your card details. There is
no third-party payment processor in this product.
We do not send your content to OpenAI, Anthropic or Google. Their usage policies
prohibit sexually explicit content, so routing this tool's drafts through them would risk
your account. How that is enforced, precisely: the direct OpenAI/Anthropic/Google keys are
hard-disabled in our configuration; the Venice models we use are Venice's own uncensored
family (Venice also resells those providers' models, which we do not select); and the
OpenRouter fallback is pointed at an uncensored open-weight model. OpenRouter and Venice are
technically capable of reaching those providers if reconfigured - our production
configuration does not do so, and doing so is prohibited by their policies.
Image generation (the Studio) is switched off in production (see §1), so no GPU host
processes your data and no face images are collected or stored today. If that changes, this
policy will be updated before the feature is enabled, and you will be told.
Each sub-processor operates under its own privacy terms. Where a sub-processor is located outside the UK, transfers are covered by the safeguards described in section 8.
To draft a reply, your fan's message text and your persona settings are sent to the AI
provider, which returns the draft text. To caption a vault image you ask us to caption,
that image is sent to the vision provider. **We do not fine-tune or train any underlying
AI model on your or your fans' data.** See §1's "Help improve Taca" and §9A for the
separate, narrower mechanism that is not model training.
Taca is built to disclose AI honestly: if a fan asks whether they're talking to an
AI, the assistant tells the truth and never claims to be human. If Fanvue requires an
AI-Creator designation on your profile for the way you use Taca, carrying that
designation is your responsibility.
We keep data while your account is active. Specific retention:
redacted after 90 days; the action record and its date are kept for your activity
totals.
they send, are cancelled, or fail - typically within minutes, never more than 24 hours.
you delete your account.
your knowledge profile, and recommendations) is kept while your account is active and
deleted with it, same as everything else here. None of it ever contains message text.
You can:
includes your per-fan notes/settings, custom folders, shot presets and credit ledger), and
vault records, per-fan notes, character-model record, audit log and any face images from
the main database, the secondary database and from disk.
Backups. Our hosting providers keep routine backups of the databases as part of their managed infrastructure. When you delete your account, your data is removed from the live databases immediately (see above); any remaining copies in provider backups are overwritten in the normal backup rotation and are not restored except to recover from a failure.
Tokens encrypted at rest (AES-256-GCM, key separate from the session secret); passwords
hashed (bcrypt); least-privilege Fanvue permissions; HTTPS enforced in production;
security headers + CSP; brute-force throttling; session regeneration on login; an audit
log of every AI action. No system is perfectly secure; we will report a serious breach as
the law requires.
Access, rectify, erase, restrict, port, or object - and withdraw consent (if you have
enabled the Studio, withdrawing consent stops the facial-image generation/training).
You can also object to "Help improve Taca" at any time by switching it off
(Settings → Help improve Taca), which stops future use and deletes what had already
been captured for it - see §1. Use the in-app export and delete tools, or email
privacy@tacatech.net. You can complain to the
ICO (ico.org.uk).
Several sub-processors are located outside the UK/EEA - currently our US-based AI providers (OpenRouter and Venice AI; Venice is the default provider for several drafting features, not merely a fallback - see the table in §3). Where personal data is transferred outside the UK, we take steps to ensure it is protected by the safeguards required under UK GDPR, such as the UK International Data Transfer Agreement (IDTA) or standard contractual clauses. The application and the main database are hosted in the UK (London).
The Studio (see §1) would process special-category facial data. A **Data Protection Impact
Assessment must be completed before that feature is enabled**, and is recommended
before public launch in any case. It is currently switched off.
Static reference material ("Layer 1"). Separately from anything described above,
Taca draws on a fixed, operator-supplied library of reference conversational material
(172 records, not sourced from your or any creator's live Fanvue conversations) to
inform how replies are drafted. This applies the same way to every creator - it is not
a per-account trial. [REVIEW NEEDED - COUNSEL] The rights/provenance basis for a
subset of this library has not yet been finally confirmed (see the internal
LAYER1-PROVENANCE-REVIEW.md for the detailed breakdown); this feature has been
enabled ahead of that confirmation as an operational decision, tracked here rather
than left undocumented.
[REVIEW NEEDED - COUNSEL] Whether creator-side "Help improve Taca" material can
lawfully include content that may reveal special-category data under UK GDPR Article 9
(for example, content revealing sex life or sexual orientation, which adult-creator
conversations may do) has not yet been finally confirmed, and no Article 9 condition is
currently relied upon for this purpose. **Cross-creator review of this material is
active** (see §1) - enabled as an operational decision ahead of this question being
resolved, not a conclusion that it no longer applies. Fan-authored material has its own
separate, still-inactive eligibility gate regardless of this. Ordinary use of Taca to
run your own account is unaffected either way.
Adults only - you, and every subject of any uploaded image, must be 18+.
We'll update this page and its date when things change, and flag big changes in-app.
Questions: privacy@tacatech.net.