The honest version, not the legalese one.
Humane Behavior AI treats your conversations as yours by default. This page describes — in plain English — what we store, what we never store, how long we keep it, and who else touches it.
Some of what follows is enforced in code. Some is an operational commitment.
Most of what follows is enforced in code: application logic, storage rules, access controls, and logging. Some of it is an operational commitment by the people running the service rather than something a machine prevents. We have marked which is which throughout, because a privacy policy that blurs the two is worse than one that admits the difference. Where a guarantee depends on an external provider, we say so.
This service is in beta. Features described here may change; when they do, this page changes with them and we tell you.
Humane Behavior AI is operated by humanebehavior.com, based in Chennai, India.
For the purposes of India's Digital Personal Data Protection Act, 2023, the operator is the Data Fiduciary. Where the EU or UK GDPR applies, the operator is the data controller. Contact at support@humanebehavior.com. If and when the service is transferred to an incorporated entity, this section will be updated before the transfer takes effect, and you will be notified.
Adults only. 18 and over.
Humane Behavior AI is not intended for, directed to, or designed for anyone under 18. Answers are direct and unsoftened, and the subject matter is adult. You must confirm you are 18 or older to create an account.
We do not knowingly collect personal data from anyone under 18. If we learn that an account belongs to a minor, we delete the account and all associated data — conversations, memories, and any training-corpus rows — and we do not retain a record beyond what is needed to prevent re-registration. If you believe a minor has created an account, email support@humanebehavior.com and we will act on it immediately.
We do not profile users by age, serve advertising, or perform behavioural tracking for advertising purposes at all — for any user, of any age.
Your chats are not used to train the model unless you opt in.
We will never use a regular user's conversations to train, fine-tune, distill, or evaluate Humane Behavior AI without your explicit, revocable permission. The default for every new account is off. You turn it on — and off — at /settings/memory.
Consent is collected by a separate, unticked control. Using the site does not constitute consent to training-data use, and continuing to use the site after we change this policy does not either. If we ever want to use your data for a materially different purpose, we will ask again.
Anonymous visitors (no account) are never eligible, regardless of any toggle. Anonymous browsing cannot produce the informed consent we would need.
One honest limitation. If you opt in and your conversations are used to fine-tune a model, we can delete the stored rows on request. We cannot remove the influence those rows have already had on model weights that were trained before your request — no one currently can. Deletion means deletion of your data from our stores, not reversal of completed training.
What we remember about you is narrow, durable, and never sensitive.
If you turn on Memory, we extract a short list of durable facts from your conversations — your name, the projects you work on, how you like to be addressed — and re-use them so later answers are less amnesiac.
We do not store anything classified as sensitive: health, finance, legal, crisis, relationships, sexuality, identity, religion. Each conversation is re-scanned before any write to enforce this even when our classifier was uncertain at retrieval time. (Enforced in code. Classifiers are imperfect: if something sensitive is ever stored, delete it at /settings/memory or tell us and we will remove it and investigate why it got through.)
You can view Memory, delete individual entries, or wipe everything at /settings/memory. Memory is independent of training-data consent — turning Memory on does not opt you into training-data use.
Handled differently, and never retained for our benefit.
If a conversation is classified as involving crisis, self-harm, or acute distress:
- It is never written to Memory, regardless of your Memory setting.
- It is never eligible for the training corpus, regardless of your training-data consent.
- No branding, share affordance, or shareable card is generated for that exchange.
It remains in your own chat history, visible only to you, and you can delete it like any other conversation.
What we do not do: we do not monitor conversations in real time, we do not alert anyone on your behalf, and we do not contact emergency services for you. The model may surface crisis resources within its answer; that is generated text, not a human intervention. If you are in danger, please contact emergency services or a crisis line directly. In India, Tele-MANAS is available free, 24/7, on 14416.
Your conversation history lives in your account, not in our training corpus.
Signed-in users get conversations stored server-side so they sync across devices. That storage is for your use — re-reading past threads, exporting them, sharing them. Saving a conversation to your history does not place it in the training corpus; those are two separate stores governed by two separate decisions.
Anonymous visitors keep conversations only in their browser's local storage. We don't see them. On first sign-in we offer to import them; if you decline, they stay browser-local.
Nothing is kept indefinitely by default.
| What | How long |
|---|---|
| Conversations (signed-in) | Until you delete them, or 3 months after your last sign-in |
| Selective memory entries | Until you delete them, or 3 months after your last sign-in |
| Account details | Duration of the account, then 30 days after closure |
| Training-corpus rows (opted-in only) | Until you withdraw and request purge, or 1 month |
| Security and access logs | 12 months, then deleted |
| Support correspondence | 24 months |
| Anonymous browser-local conversations | We never receive these; they live only on your device |
If your account is inactive for 12 months we notify you at your registered email before deleting anything. When you delete something, it is removed from live systems immediately and purged from backups within 30 days.
No public sharing without your explicit click and review.
We do not produce public links from your conversations automatically. Any public-share feature requires you to click “Create public share link,” review the exact text that will be published, and confirm. Personal data — email, name, attachment filenames, and selective memory context — is stripped from the published copy before the link goes live.
You can revoke a share link at any time; revocation takes effect immediately, though we cannot recall copies someone already made. Crisis-classified exchanges cannot be shared publicly.
We do not sell, rent, or share your data with third parties for their own use.
Your conversations, memories, uploaded files, and account information are never sold, rented, or shared with advertisers, data brokers, marketers, or any other party for their own purposes. We do not run ads. We do not build advertising profiles. We do not license your data to model providers for their training.
The only third parties involved in operating the site are inference-API providers (the LLM APIs that generate answers) and infrastructure providers (hosting, database, email delivery). Each processes your data only as needed to deliver the service, under contractual confidentiality, and is contractually prohibited from retaining it for their own model training. We do not send them your identity beyond what the prompt itself contains.
Cross-border transfer. Because these providers operate outside India, your data is processed outside India. We rely on contractual safeguards with each provider. Where EU/UK transfers are made, they are made under Standard Contractual Clauses or an adequacy decision.
Lawful demands. The only other exception is a lawful, narrowly-scoped request from a competent authority that we cannot legally refuse — in which case we respond only to the specific data compelled, and we notify you unless the request legally prohibits notice.
If something goes wrong, you hear it from us.
We protect your data with encryption in transit and at rest, access controls limiting who can reach production data, and audit logging of administrative access. (Enforced in code and infrastructure configuration.) No system is perfectly secure, and we are not going to claim otherwise.
If a personal data breach occurs, we will:
- notify affected users without delay, describing what happened, what data was involved, and what you should do;
- notify the Data Protection Board of India within 72 hours, and any other regulator where legally required;
- publish what we found and what we changed.
Report a suspected breach to support@humanebehavior.com.
Only strictly-necessary cookies.
Keeping you signed in and protecting the sign-in handshake. No advertising, no third-party trackers, no cross-site tracking.
We log first-party pageviews and clicks so we can improve the product — no third-party trackers, no advertising.
To understand how the product is used and diagnose problems, we record pageviews, click positions on buttons and links only(never on inputs or text areas), your device / browser / OS, and an approximate geo location derived from your IP. Values you type — chats, prompts, form fields — are never captured.
Geo lookup: we send only your IP address to ip-api.com to resolve it to a city and country, then cache the result. No other data is shared. If you prefer even this lookup not happen, write to support@humanebehavior.com and we'll suppress geo for your account.
Retention / product emails: we may occasionally email you from support@humanebehavior.com about your usage (e.g., a note if you're close to your monthly cap) using Resend as our email provider. These are transactional, not marketing.
Analytics data is stored on our own servers and used only by our team. You may write to support@humanebehavior.com to request deletion of your session history.
The model generates text. It does not decide anything about you.
It does not make decisions that produce legal or similarly significant effects about you — no scoring, ranking, eligibility determination, or profiling that feeds a decision about you. Everything it produces is output for you to read and judge.
Every answer is a probabilistic guess. Verify before you rely on it.
The AI can produce answers that are factually wrong, ethically questionable, or entirely invented — sometimes in a very confident tone. Names, citations, statistics, and quotes may not correspond to anything real. Do not use the output as medical, psychological, legal, financial, or safety advice. Always verify before you act. See the Terms of Use for the full disclaimer and liability terms.
Access, correction, erasure, withdrawal, nomination, and complaint.
Right to access. You can see every conversation, every stored memory, and every account detail we hold, from inside the app.
Right to correction. If anything we hold about you is inaccurate, incomplete, or out of date — including a Memory entry — you can correct it in the app or ask us to.
Right to erasure. You can delete any conversation, any memory, or your entire account. Full account erasure includes historical training-corpus rows on request. Subject to the limitation noted under Training data regarding model weights.
Right to withdraw consent. Turn the training-data toggle off at /settings/memory at any time. Withdrawal is as easy as giving consent. From that moment no new rows enter the training corpus; email us to also purge rows saved while consent was on.
Right to nominate. Under the DPDP Act you may nominate someone to exercise these rights on your behalf in the event of your death or incapacity. Email support@humanebehavior.com to register a nominee.
Right to complain. See below.
Every privacy concern is reviewed by a named human, the same business day.
If something looks wrong — data appearing where it shouldn't, a consent setting not being respected, an unexpected use of your conversations — email support@humanebehavior.com. A human reviews it the same business day and responds with what happened and what we are doing about it. We aim to resolve within 30 days.
Internally, the service is designed so violations of the rules above are hard to introduce accidentally and detectable when they occur. We keep an audit trail so any concern can be traced, explained, and corrected.
This policy is published in English.
If you would prefer it in Tamil, Hindi, or another language in the Eighth Schedule to the Constitution of India, email support@humanebehavior.com and we will provide it.
We update this page as the product evolves and as the law changes.
Material changes are flagged at the top and notified by email to registered users. Continued use is not consent. Where a change materially alters what we do with your personal data, we ask for fresh consent rather than assuming it from your silence. If you do not agree, you can close your account and take your data with you.
Questions about this policy? support@humanebehavior.com