Revision date: 5 October 2026
This revised policy takes effect when published at into3.ai/privacy.html.
Into3 is operated by DEEPSCIENCE COGNITECH AI LABS PRIVATE LIMITED (“Into3”, “we”, “us” or “our”), at A-116, First Floor, Urbtech Trade Centre, Sector 132, Noida, Gautam Buddha Nagar, Uttar Pradesh 201304, India.
This policy covers into3.ai, campaign and reservation pages, the RTAE demo, and learning services that link to it. Into3 is being developed for students in Classes 6–12, primarily in India. Features may become available at different times; a feature described here is not necessarily included in every service or session.
“Parent” includes a person legally authorised to act as a child’s guardian. For our Indian child-account process, a child is a person under 18. Adult learners exercise their own privacy choices. Personal information includes identifying details, information reasonably linkable to an individual, device identifiers and identifiable inferences.
Relevant Into3 features use camera and microphone inputs. Our public website and reservation journey use Google Analytics and Meta advertising measurement. These activities are explained separately below. Booking access, paying ₹100 or reading this policy does not provide consent to collect a child’s camera, microphone or physiological information, or to use their information for model retraining.
The choices, safeguards and retention limits in this policy apply from its effective date. Required feature-specific notices and consent requests are provided before the relevant processing starts.
The reservation form collects the parent’s name, email address, mobile number, state or union territory, and the student’s class. We also process parent declarations, accepted terms, reservation references, payment and refund status, and support correspondence.
We use these details to manage reservations, communicate about access and launch, reconcile payments, handle support and refunds, prevent misuse and maintain required business records. Pre-booking does not require a child’s camera or microphone input. Optional promotional messages require a separate choice; necessary booking and account messages are not optional marketing.
Razorpay processes payments through its payment interface. Into3 receives transaction details such as the order or payment identifier, amount, currency and status. Our reservation form does not collect full card numbers, card security codes, UPI PINs or banking passwords. Please do not send these to us. Section 10 explains measurement associated with the booking journey.
We process information relevant to the selected service, such as the learner’s name or account identifier, age or class, curriculum, language, interests, goals, preferences, authentication details and session records.
During a lesson, we process questions, answers, messages, spoken responses and submitted notebook work. PRIA Dock may capture the notebook area to check working steps. Images can incidentally include names, hands or nearby objects; keep unrelated personal material outside the camera’s view.
We distinguish original inputs from structured learning records: concepts attempted, extracted working-step summaries, scores, mistakes, hints, response times, practice results, mastery estimates and learning preferences. The Individual Learning Planner, or ILP, uses these records to maintain continuity and select language, tone, depth, pace, examples and reinforcement. Familiar-context examples use information supplied through the service; this does not itself involve tracking precise physical location.
Original learning-session inputs are handled under Section 5. Structured records have the limits in Section 13. A record is not anonymous merely because it uses an account code instead of a name.
Blob’s World conversations are processed to respond within the conversation. Verbatim conversations and transcripts are not retained as an archive, added to academic assessment or included in general model-training datasets. Live conversation context is cleared under the raw-input rules below. This is not a medical service, and conversations do not carry medical professional privilege.
The Real-Time Adaptation Engine, or RTAE, combines learning responses with available observations to adapt teaching. Depending on the feature and permissions, it processes:
RTAE uses a 47-parameter framework. The feature notice identifies the inputs and estimates used in that session; every parameter is not necessarily collected in every session or directly measured by a camera. We explain additional physiological processing and its source before enabling it.
Physiological and emotional estimates can be sensitive personal information. They may be inaccurate and are not direct proof of a child’s thoughts, understanding or health. They support teaching decisions, such as changing an explanation, pace or suggested break; they are not medical diagnoses or substitutes for clinical measurements.
We do not save raw learning-session inputs as retained recordings or content archives. This covers original camera frames, microphone audio, notebook images, verbatim spoken or typed answers, transcripts and conversation text. It does not mean no information is retained: parent/account details, derived RTAE measurements and structured learning records have separate purposes and limits.
Images and audio are held temporarily for the relevant processing operation and discarded when it finishes, is cancelled or fails. Text needed for an ongoing conversation remains in temporary session memory until the session ends or reaches 30 minutes of inactivity, whichever occurs first. Raw content is excluded from routine debug samples, persistent application logs, backups and general model-training datasets.
Analysis can happen on the device or on Into3’s backend. The public RTAE demo sends camera frames to an Into3-operated backend hosted on Heroku. In this configuration, raw frames leave the device for analysis; transmission is not limited to numbers. Transmission and storage are different activities.
Some voice interactions use browser speech recognition. Depending on the browser, its provider may process audio remotely. Embedded YouTube lessons also involve Google/YouTube processing. Before enabling such a feature, we identify its processing route and obtain the notice and authorisation required for it. Providers’ own processing is explained in their notices; Into3’s no-storage rule describes our handling and the restrictions we require from processors acting for us.
Browser camera and microphone access requires device permission. You can decline or revoke it and stop the relevant session. Browser permission does not replace required parental consent. Revoking it stops further access through that permission but does not delete previously derived records; deletion can be requested separately.
We use information for the disclosed activity:
Automated systems generate explanations, recommendations and learner-state estimates. You may question an output or ask us to correct an inaccurate personal record or inference. A learning profile is not a clinical assessment or a conclusive statement of ability.
Before collecting a child’s personal information for learning or sensing, we give the parent understandable information about the data, purposes, recipients, retention and choices. We obtain and verify parental consent where required. Payment, a device permission or an unverified claim to be an adult does not replace the required verification.
We collect only what is reasonably necessary for the disclosed activity. We separate ordinary learning, optional sensing, model retraining, research and marketing choices. A child’s access is not conditioned on giving unnecessary information. Parents can decline optional sensing or model retraining without losing access to learning features that do not depend on that input.
You can withdraw consent through a control provided with the feature or by emailing info@into3.ai. We explain any feature affected by withdrawal. Withdrawal stops the relevant future processing; it does not invalidate earlier lawful processing or remove a specific legal record-retention duty.
Consent does not override legal restrictions on child monitoring or advertising. If you believe information was collected without the required authorisation, contact us. We will investigate, restrict inappropriate further use and address deletion or other corrective action.
Into3 uses its own learning model and uses data to improve and retrain it. We do not permit external providers to use learner inputs to train their own models. Hosting our model with an infrastructure provider is different from authorising that provider to train on learner information.
Under this policy, general model retraining is limited to separately authorised derived RTAE features and structured learning records, such as response timing, accuracy, error categories, mastery estimates and the result of an explanation or intervention. Original camera/audio/notebook inputs, verbatim transcripts or chats, Blob conversation content, payment information and direct contact details are excluded.
We obtain a separate affirmative parent choice before including a child’s eligible records in retraining. This is optional and is not a condition of ordinary learning. Authorised records may be used in a later training batch. Personalising the current lesson is a separate service purpose and does not by itself authorise general retraining.
Training records use coded identifiers with direct identifiers removed. A restricted mapping may remain to administer consent and deletion, so they are treated as personal information, not automatically as anonymous data. Access is limited to authorised model-development personnel and administrators who need it. Section 13 applies to every copy and derived training file.
Parents can stop future inclusion by contacting our privacy team. We exclude the child’s records from future batches once the request and authority are verified and remove eligible training records within 30 days, or sooner where required. Deleting source data does not automatically reverse a completed model update. We assess whether a retained model contains identifiable personal information and address applicable rights through appropriate removal, restriction or model-remediation measures; this distinction does not remove those rights.
Separate research studies require their own prior notice and authorisation, including the study’s data, recipients, end date, retention and withdrawal arrangements. This general policy does not authorise saving raw inputs for research or publishing identifiable examples. A study proposing this requires a separate explanation and consent before collection.
Into3 operates the learning and RTAE services and remains the contact for questions about its processing. These services have distinct roles:
We also use service providers for necessary hosting, database/storage administration, email delivery and support. They receive only information needed for their assigned work. If you choose WhatsApp or another external communication service, its provider also processes that communication under its own notice.
Processors acting for Into3 are subject to instructions covering purpose, access, confidentiality, security, deletion and restrictions on their own use. We assess new providers before giving access. Before an additional operator collects or maintains children’s information, we identify it through the applicable updated policy and direct notice and obtain renewed consent where required. Contact Into3 for provider and location information relevant to your service.
Internal access is limited to authorised people who need it for service delivery, support, security or a legal obligation. A school or educator receives learner information only through a disclosed and authorised connection. We may disclose information when lawfully required, to address fraud or security incidents, or during a corporate transaction subject to applicable safeguards and notice.
We do not sell children’s learning, camera, voice or physiological information. Adult website disclosures to measurement providers are described separately below.
Our public website and reservation journey use cookies, browser storage and server requests. Necessary storage supports reservation state and security. Optional analytics and advertising measurement require a prior choice. You can decline both; agreeing to analytics does not automatically agree to advertising measurement.
Google Analytics measures visits, interactions and browser/device information. Under this policy, user-level and event-level retention is set to two months, with reset on new user activity disabled; Google applies deletion through its processing cycle. Standard aggregated reporting is not subject to that same setting. Google Signals and Google advertising-personalisation features are disabled. We do not send names, emails, phone numbers, child learning content or physiological information to Analytics. Google explains Analytics retention here.
When advertising measurement is accepted, Meta Pixel and related server-side events measure visits and configured reservation/payment events. Data may include page address, event type/time, event identifiers, browser/network information, advertising click identifiers and cookie identifiers. Matching event identifiers help avoid counting the same event twice.
Where a parent accepts advertising measurement, the booking journey may hash their email and phone number for event matching. Hashes remain personal information because they can be matched to a person. Meta may combine received events with its own information under its terms and controls. Consent applies to browser events and server-side forwarding.
Our storage limits are:
You can refuse optional measurement at the consent prompt and change your choice through the website’s consent controls. You can also contact info@into3.ai about withdrawal or deletion. Browser cookie/site-storage controls and provider settings offer additional controls. Clearing local storage does not itself delete server records; blocking cookies alone may not stop server requests.
We stop affected optional collection and forwarding after withdrawal and address identifiable information already held through the applicable deletion process. Into3’s identifiable campaign-event records have a maximum 90-day retention from the event. Security and financial records have separate purposes and periods; provider-held information is subject to applicable provider arrangements and rights.
Advertising and marketing tags are excluded from child learning, RTAE sensing and companion sessions. Child names, schoolwork, conversations, physiological values and inferred emotional states are excluded from advertising events, URLs and audiences. A marketing label alone does not establish that a visitor is an adult.
Information is processed on the user’s device, in Into3-operated systems and by relevant providers. AWS, Heroku/Salesforce, Google and Meta operate internationally. Processing, support or supporting infrastructure may be outside India; we do not promise that every copy or processing operation stays in India.
We assess the destination, recipient, purpose and protection before transferring learner or physiological information. We require purpose-limited processor arrangements, appropriate access restrictions and security protections, and any consent or transfer mechanism required by applicable law. A processor may not add an unapproved location or subprocessor contrary to those arrangements.
Where a transfer notice or permission is required, the feature notice identifies the recipient and destination before transfer. A provider’s general availability of an Indian region does not establish our configuration. Contact info@into3.ai for location and safeguard information for your service. We comply with applicable government transfer restrictions and local retention requirements.
Our safeguards include encrypted transport, appropriate protection for stored personal information, access limited by role and need, and controls over providers, logs and backups. We do not claim a particular security certification or that all processing is unreadable to Into3: a server analysing content can access the information needed for that analysis.
No system is completely secure. If a breach occurs, we investigate, contain and remedy it and provide notifications required by applicable law within the relevant deadlines.
These limits govern ordinary service use. We delete information earlier when its purpose ends or a valid request requires it. Specific statutory retention and documented legal holds take precedence where applicable. Legal copies are restricted to that purpose and are not reused for teaching, training or marketing merely because they remain stored.
No persistent recordings or content archives. Camera frames, audio and notebook images are discarded after the processing operation. Verbatim answers, transcripts and chats needed as live context are cleared at session end or after 30 minutes of inactivity, whichever occurs first. These inputs are excluded from routine logs, debug samples, backups and general training datasets.
Maximum 30 days from collection, including copies used in authorised analysis or training. Copying does not extend expiry. Separate learning summaries, such as a concept requiring revision, follow the learning-history limit; they do not preserve a detailed physiological or emotional timeline.
Kept while needed for an actively used learner account. Eligible records are deleted within 30 days of a verified deletion or closure request. After 12 months without learner-account activity, we delete them unless the parent confirms continued use. We send notice at least 30 days before inactivity deletion. This category does not include notebook-image archives or verbatim Blob conversations.
Authorised structured training records, derived files and annotations are kept for at most 90 days from original collection, or until the training purpose ends, whichever is earlier. Detailed RTAE measurements retain the shorter 30-day limit. Eligible records are removed within 30 days of verified withdrawal or deletion, without extending an earlier expiry. Completed model versions may remain while used and for up to 90 days after retirement for recovery or evaluation, subject to applicable rights concerning any personal information they retain. We do not assume model updates are anonymous.
A study has a fixed end date and a specific retention limit disclosed before participation. The default limit is 90 days from collection, with earlier deletion when the purpose ends or a valid withdrawal requires it. A different necessary period or raw-input recording requires an explicit study-specific notice and authorisation before collection; the general service policy does not grant it.
Unpaid reservations are deleted 90 days after creation unless completed. Necessary non-financial records of completed or cancelled reservations are kept until fulfilment or closure plus 90 days. Support correspondence is kept for up to 12 months after resolution. Optional marketing contacts are removed on withdrawal or after 12 months without relevant engagement. Required transaction or complaint evidence is separated from these records.
Financial books and supporting transaction/refund records are retained for eight financial years following the financial year to which they relate, or longer for a specific legal requirement or investigation. This does not extend retention of unrelated child learning or physiological information. Payment providers also have their own legal obligations.
Minimum evidence of the choice, notice version, verification, dates and action taken is retained while the relevant authorised processing continues, and for one year after the later of its end or resolution of the related request. A specific legal duty or active dispute can require longer restricted retention. Raw child recordings are not kept simply to prove consent.
Applicable ICT security logs are kept for a rolling 180 days within India under the relevant CERT-In requirements. A longer applicable legal period governs the particular logs or processing records within its scope. Raw learning content is excluded from routine logging. Legal holds identify the records, reason and review date.
Eligible information is removed from active systems within 30 days after a deletion request is verified, unless an earlier deadline applies. Ordinary backup copies expire within a further 30 days after active-system deletion. They are isolated from ordinary use, and a restoration must reapply deletion instructions. The normal maximum is 60 days from verified request to backup expiry, subject to specific mandatory retention or a documented legal hold.
Section 10 covers website storage. Provider instructions require retention and deletion consistent with the relevant purpose. If a legal exception prevents deletion, we explain the retained category, reason and applicable period unless law prevents that explanation.
As additional data-protection retention duties come into force, they apply to records within their scope. We update the relevant notice and restrict retained copies to their legal purpose. Retention is not permission for an unrelated use. Coding or hashing alone does not make a record anonymous.
Contact info@into3.ai to ask about information held about you or your child, seek access or correction, request deletion, withdraw consent or exercise another applicable right. Parents may stop further collection or use. Depending on the applicable law, rights may also include restriction, portability, nomination and regulatory complaints.
Identify the service and provide enough account or reservation information to locate the record. Do not email passwords, payment PINs, full card details, raw recordings or identity documents. If verification is needed, we explain the proportionate information and secure method required. We check parental authority before disclosing a child’s information.
We respond within the applicable legal deadline and explain any lawful refusal, required retention or extension. Indian grievances governed by the Sensitive Personal Data or Information Rules are addressed expeditiously and within one month of receipt. Verification is not used to delay applicable response deadlines.
Device permissions, reservations, learner accounts and training consent are separate. Tell us your request’s scope and we will help identify the corresponding processing. Withdrawing optional marketing does not cancel necessary account or order communications.
We apply requirements relevant to the activity and individual. In India, these include applicable provisions of the Information Technology Act and Sensitive Personal Data or Information Rules, and the Digital Personal Data Protection Act, 2023 and its rules as their provisions come into force.
Where US children’s privacy law applies, COPPA protects children under 13 and requires the applicable notice, consent, parental rights and other safeguards. Into3 has applied to a certification company for COPPA-related certification; the application is pending. This is not an issued certificate or an approval or endorsement by the US Federal Trade Commission.
If another jurisdiction’s law applies, we meet the relevant additional notice, lawful-basis, children’s protection, rights and transfer requirements before the corresponding processing. This policy does not represent that every country is a supported market. Contact us about the country and service concerned.
Links and integrations with WhatsApp, YouTube, Razorpay and other services involve their own notices. Embedded content can send information when loaded, before you leave Into3. We remain responsible for obligations applying to our selection and use of integrations; a provider-policy link does not remove them.
We update the revision date and give appropriate notice of material changes. Where a new direct notice or renewed consent is required, particularly for children’s information, we complete it before starting new processing. Continued use is not automatically consent to a materially different purpose.
A correction does not retrospectively obtain consent for earlier collection. Previous versions may be requested from our privacy contact. Each revised version’s publication date is recorded in our policy history.
Grievance Officer: Bhupesh Sarangal.
Email for all concerns, privacy requests and grievances: info@into3.ai.
Operator: DEEPSCIENCE COGNITECH AI LABS PRIVATE LIMITED, operating as Into3.
Address: A-116, First Floor, Urbtech Trade Centre, Sector 132, Noida, Gautam Buddha Nagar, Uttar Pradesh 201304, India.
If a concern remains unresolved, you may use the complaint or escalation route available under applicable law. We will provide information about that route on request.