HELP AI Transparency Notice

1. When you interact with AI

HELP uses AI to understand requests, ask useful follow-up questions, organize information, explain possible matches, support search, draft or translate text, transcribe voice and generate spoken replies where the relevant functions are enabled. HELP identifies AI interactions so you can distinguish a model from a human participant or HELP team member. HIC and AI CORE are AI-supported features, not human operators.

Some recommendations come from structured rules, semantic search and matching algorithms without a conversational model. HELP does not populate real search or preliminary matching with fabricated people or simulated offers. Non-personal instructional examples, when used, are distinct from an actual candidate and cannot create a person's consent or commitment. Applicable AI-generated or manipulated-content marking is provided for supported output types as required by law.

Part of the scenes and the music of HELP's film are created with AI, and a line under the film says so. HELP's avatars "She" and "He" are fictional people whose faces and movements were made with AI, not real people. Images of them that leave HELP carry a "Created with AI" mark. Their lip movements may not yet exactly follow the words.

Models and component licences. HELP may use proprietary and open-source models, including copyleft-licensed server components, only after checking the actual licence for commercial services, distribution, modification, attribution and any source-disclosure obligations. New model deployment does not excuse provider, privacy or licence review.

2. What AI can and cannot decide

An AI response may contain mistakes, omit context or misunderstand what you intend. Check descriptions, locations, times, prices, health or safety information, translations and agreements before using them. AI output is not professional legal, medical or financial advice and is not an emergency response.

AI does not independently sign an agreement, spend your money without the applicable authorization, approve a statutory refund, decide a serious suspension or resolve a formal moderation appeal. Important matters requiring human authority follow the applicable approval or review route. Temporary automatic blocking of an unsupported or unsafe action is not itself a final finding of wrongdoing.

The words of a reply appear as they are produced. If HELP's checks find an error, the reply is collapsed and replaced. Replies with money amounts, consent or approvals and legal confirmations appear only after checking.

When you ask HELP to do something, the AI only proposes the action. HELP checks the proposal against your settings, permissions and HELP's rules and carries it out only through HELP's regular functions, with the same checks as when you use them yourself; the AI does not carry out actions on its own. HELP asks you to confirm before an action in which money leaves you (a purchase, a subscription, a gift, a donation, a payment to another person or a Connection fee), before an action that cannot be undone, and before an action that shares your personal information with another person or changes who can see it. You confirm such actions by tapping or clicking, not by voice. Charges for the AI work itself are described in Section 4.

AI-generated suggestions do not disclose a match's identity or contact another person without the required privacy and action controls. You may correct, reject or ask to review an output. Ordinary search or other available non-AI paths can remain available when AI functions are not enabled.

2A. Actual search, preliminary matching and empty results

Eligible participants may use ordinary and AI-assisted discovery under the same baseline network permissions, subject to credit authorization where AI operations are charged, disclosure settings, safety restrictions and market availability. No restricted HELPER tier or fictional person is required as an intermediate experience. HELP must not display synthetic demonstration profiles as actual users, offers or available professionals.

A preliminary compatibility calculation does not reveal a person's identity or guarantee that the candidate can be contacted. Zero returned results may reflect available data, visibility, filters, timing or system limitations. HELP may show an appropriate explanation and an optional invitation/share action; neither the model nor the platform guarantees that a desired person exists or will respond. AI must not represent an unverified request for money or credits as a verified need.

3. Your information and provider processing

A requested AI function may receive what you provide, limited relevant personal context permitted for that function, structured fields derived from it and its resulting output. Voice transcription involves microphone/audio content when you initiate that action; voice synthesis may transmit approved text to a speech provider. For voice, HELP uses Amazon Web Services: Amazon Transcribe to recognize speech and Amazon Polly for spoken replies. Never record another person unlawfully.

One-off phone calls are available in the countries where our telephony provider supports them, and only after the user's «yes» to that specific call. Where HELP offers to make a one-off phone call for a user's approved case, the AI clearly identifies itself as AI acting on the user's behalf at the opening of the call. The user is informed in advance that calls may be needed within that case, their permitted purpose, relevant personal details to be disclosed, and credit-spending rules. A further separate approval for each routine permitted call is not required if that call remains within the specific approved case and authorization; a materially expanded purpose, financial commitment or new use of personal data requires an appropriate additional authorization. The AI may not agree to make a separate payment or create a commitment outside the user's authority. Sound recording is off by default; any transcript, summary, and notice to the called party are governed by applicable law and the Privacy Policy. Calls are enabled only for country-and-purpose combinations cleared under applicable calling and data-protection rules. A party's do-not-call request is respected. Charges are subject to the approved plan/credit limits and never generate automatic card charges.

AI providers and regions differ by activated route. We do not promise that every AI call remains in Frankfurt or that every outside provider deletes all information immediately. The Privacy Policy explains legal bases, actual recipients, regions, retention and international transfers. Certain external facts — such as place details, map data or store prices — can be outdated and subject to provider licences; where applicable, we identify the source and time checked. HELP is not automatically the seller or affiliate of a linked shop.

Personal AI memory for your own use is different from encrypted operational quality records and from voluntary use of information to improve shared HELP systems. Shared learning is off by default and must have a separate, freely chosen, revocable permission before eligible personal data may be used for that purpose. Your consent does not authorize unrelated external provider training or training on another person's private content.

Facts from external sources

Where the relevant lookup is enabled, HELP may show information from open map sources and licensed outside suppliers about places, opening hours, routes or store prices. Such information can change and should be treated as checked at the displayed time, not as a guarantee of current availability, the lowest price or exact travel time. We display required attribution, for example © OpenStreetMap contributors, and provider names or marks where contractually required. HELP does not receive a store affiliate commission or sell organic ranking through these lookups in the approved V1 model. Licensed third-party retrieved information is not automatically available for training HELP's own AI.

Photo-based personal avatars (where enabled). In V1, personal avatars based on a real face are limited to adults aged 18 or older using their own image. Before processing, HELP obtains a separate explicit, purpose-specific consent, discloses involved providers, relevant international transfers, storage/deletion rules and available withdrawal controls. Using another person's face or a child's face is prohibited. No voice cloning is offered: eligible avatars use licensed standard synthetic voices. Where required, synthetic or manipulated media is labelled visibly and with suitable machine-readable provenance; the person may report misuse. HELP stops further consent-based use and erases controlled active copies upon an applicable deletion request, subject to backup cycles, providers' verified obligations and legal exceptions. Commercial advertising featuring a user's likeness requires separate permission. A provider may process images only after its contracts, locations, retention and commercial use rights have been cleared. Biometric identification, if used, requires its own Article 9 GDPR assessment and appropriate safeguards; consent alone is not a blanket approval to deploy any process.

HELP film and synthetic character rights. AI-generated or manipulated publicity content is disclosed in publication descriptions, technically labelled where required, and marked in the video itself when necessary under applicable rules. The synthetic 'She' identity is not cleared for commercial use until its provenance, commercial rights and any recognizable real-person likeness risk have been reviewed. HELP's internal choice not to launch a promotional avatar fund creates no user entitlement to free funded videos.

4. Paid AI operations

Eligible AI, search, voice transcription, spoken replies and translation are paid from your plan's monthly limit and then from AI Credits under the default consent you give by accepting the Terms of Service and the Plans and AI Credits Terms (PUB-07). HELP does not ask you to confirm an amount before each step; every debit is shown in your history with what was done, when, how much and why. Before a paid step HELP holds no more than that step's maximum; only the actual cost is debited and the rest is released. If your plan limit and credits are not enough, the work does not start.

HELP currently selects the model, provider route and internal effort automatically according to applicable quality, qualification, safety, availability and cost policies. HELP must not use an unqualified model to lower the cost. User-selectable model effort is not currently offered; any later option would require separate disclosure. Provider pricing changes do not automatically alter HELP's published rates or charges already settled. AI can make mistakes and give inaccurate or incomplete results; AI work is paid even when a step does not fully reach the result. If you stop a step or its result is below the required level, only what was actually spent is debited, with the reason in your history. An AI error the provider charged for is debited at cost, with no markup; a provider failure (it charged nothing or did not work technically) or a HELP failure costs nothing (PUB-07, "What you pay for").

Default consent does not cover new purchases of plans or credits, card payments, gifts, donations, payments to other people or the fixed Connection fee; each needs its own confirmation. For paid transcription, voice synthesis or translation, the participant who asks pays from their own plan limit or credits. Asynchronous human support is free; AI in support is paid like other AI work. Paid features depend on product and market availability. See the Plans and AI Credits Terms and Consumer Purchase Terms.

5. Transparency, feedback and human review

AI should be identifiable at or before your interaction as required by law. Where generated audio, images, video or other output needs a special label or machine-readable marker, that requirement is handled by the relevant function; this Notice alone does not provide such a marker. Do not falsely present synthetic output as authenticated human testimony.

You can report an error or harmful answer and request appropriate human review via HELP support. A human response may take time; there is no round-the-clock human chat or promise of immediate help. Statutory rights and deadlines remain in effect.

HELP will assess separately any future use of AI in professional selection, work allocation, financial entitlement, essential services or other potentially high-impact decisions. We do not declare every possible HELP AI use low-risk by default.

5A. HELP Eyes, external research and user-assistant estimates (V5.2)

Estimates of prices, locations, journey times, availability and external source descriptions produced by the user's own assistant or HELP Eyes may be inaccurate or out of date. They are research aids, not verified facts, price offers, guaranteed routes or professional advice. Where available, show the source and retrieval time and let the user review them. An AI-generated estimate must not be presented as a real person's commitment or final agreement term without independent confirmation.

5B. User guidance and real-world actions

HELP may explain how to use the service through instructional text and non-personal interface examples. Such guidance is not a real person's offer and does not create a Help Agreement or payment. Any real help relationship, purchase or AI charge still requires its own applicable confirmation. A fictional-person training journey is not part of the general participant discovery experience.