Contents — 14 sections
- Scope and definitions
- 1. You are talking to software
- 2. The output is generated, not found
- 3. What generated text can get wrong
- 4. Your responsibility before you publish
- 5. Two things we would rather tell you than let you discover
- 6. Which AI providers we use, and for what
- 7. How your style profile works
- 8. OPEN ITEM — machine-readable marking of AI-generated text
- 9. Your own obligations when you publish
- 10. What we do not claim
- 11. Changes to this disclosure
- 12. Contact
- Contact
Scope and definitions
In this disclosure, "we", "us" and "our" mean IX Labs. "You" and "your" mean the person using our product.
Our product is Ingy, a Telegram bot that drafts social media posts from a link, a video, or plain instructions. Our website is ingy.app and the bot is @IngyAppBot.
This disclosure stands on its own. You do not need to read any other document to understand what is written here.
1. You are talking to software
1.1 Ingy is an artificial intelligence system. When you send a message to @IngyAppBot, you are interacting with software. You are not talking to a person.
1.2 No human at IX Labs reads your messages as part of drafting your posts. The drafting is done by machine.
1.3 We are telling you this plainly because we think you should know it, and because EU law now requires providers of AI systems that interact with people to make it clear that the interaction is with an AI system.
2. The output is generated, not found
2.1 Ingy does not search a library of existing posts and hand you one. It does not retrieve, quote or copy an answer.
2.2 A large language model composes new text, word by word, based on the source you gave it, your instructions, and your style profile. Every draft is newly written.
2.3 This matters for two reasons that most people do not expect, and we set them out in sections 3 and 5.
3. What generated text can get wrong
Read this section before you publish anything Ingy writes.
3.1 Output may be inaccurate. It may state something that is simply not true.
3.2 Output may be incomplete. It may leave out the part of the source that mattered most.
3.3 Output may be out of date. The model was trained on information up to a point in time and does not know what happened after it.
3.4 Output may be entirely fabricated while appearing completely plausible. This is the failure mode people miss. A generated post can invent a statistic, a quotation, a date, a study, a product name or an event, and present it in confident, fluent, well-formatted language that reads exactly like a fact. Fluency is not evidence of accuracy.
3.5 Output may misstate facts about real people, places, organisations and events, including attributing statements or actions to people who never said or did them.
3.6 These are known characteristics of the technology. They are not bugs we expect to eliminate, and no setting in Ingy switches them off.
4. Your responsibility before you publish
4.1 You must review and fact-check every draft before you publish it. Check names, numbers, dates, quotations, claims about people and organisations, and anything that would matter if it were wrong.
4.2 Ingy never publishes anything. Publishing is entirely manual. Ingy does not post to any channel, page, feed or account, and has no ability to do so. It hands you plain text in a Telegram message. Nothing leaves Ingy unless you copy it and post it yourself.
4.3 When you post that text, you are the publisher of it. You are responsible for it in the same way you would be responsible for anything else you write and publish, including under defamation, advertising, intellectual property and consumer law.
4.4 Output does not represent the views of IX Labs. We do not review, approve, endorse or adopt anything Ingy drafts for you. Text generated for you is your text and your decision.
5. Two things we would rather tell you than let you discover
We think an honest disclosure is worth more than a flattering one.
5.1 Similar or identical output may be generated for other users. The same or a very similar draft can be produced for someone else who gives a similar source and similar instructions. We do not and cannot promise that anything Ingy writes for you is unique to you.
5.2 Purely AI-generated content may not qualify for copyright protection in some jurisdictions. Several legal systems require human authorship for copyright to arise. Text produced by a machine, without meaningful human creative input, may not be protectable at all in those places. If ownership of the finished text matters to you, edit it yourself, add your own material, and take your own legal advice. We make no claim or promise about whether you own copyright in Ingy's output.
6. Which AI providers we use, and for what
Ingy is built on third-party AI services. We name them so you know where your content goes.
| Provider | What it does |
|---|---|
| Anthropic | Generates the text of your drafts. Receives your prompts, sources and instructions, and returns the output. |
| OpenAI (Whisper) | Transcribes audio. Where you send a voice message or a video, the audio is transcribed to text so it can be used as a source. |
6.1 Other providers support the product — fetching article pages, retrieving video transcripts, tracing model calls, analytics, hosting and payments. They are listed in full on our Sub-processors page.
6.2 We do not control how these providers build or behave. We select them, and we can change them; if we add a new provider that will have access to your content, we will say so.
7. How your style profile works
7.1 Ingy keeps one style profile per account, learned from your own finalised posts. Nothing else feeds it.
7.2 The profile covers six facets: voice, format, lexicon, metrics, rules and avoid. In plain terms: how you sound, how you structure a post, the words you use and do not use, the numbers you tend to cite, the rules you have set, and the things you have told Ingy never to do.
7.3 The profile rides along in the prompt sent to the model, so drafts move closer to how you actually write.
7.4 You can view and edit your style profile. It is yours. If it has learned something you disagree with, change it.
7.5 Your style profile is never shared with another user and is never applied to another user's posts. It is scoped to your account only.
8. OPEN ITEM — machine-readable marking of AI-generated text
This section is deliberately labelled as an open item. We are not claiming compliance here. We are telling you where we are.
8.1 Article 50(2) of the EU AI Act requires providers of AI systems that generate synthetic text to mark outputs in a machine-readable format so they are detectable as artificially generated or manipulated, subject to exceptions and to technical feasibility and state of the art.
8.2 Article 50 applies from 2 August 2026, with a transitional grace period to 2 December 2026 for AI systems already placed on the market before that date.
8.3 The practical difficulty for a product like Ingy is honest and specific. Ingy's deliverable is plain text inside a Telegram message that you copy and paste into another platform. It is not a file we deliver, not a document with metadata we control, and not a post we publish. Any machine-readable marker embedded in the text — a watermark, an invisible character, a metadata wrapper — is stripped, broken or made useless the moment the text is selected, copied and pasted, and would in most cases damage the very thing you are paying for: clean, publishable copy.
8.4 We are assessing how the marking requirement applies to this delivery model, including what the state of the art in text watermarking actually supports, which exceptions apply, and what our AI providers implement at their end.
8.5 We commit to updating this page when we reach a position, and to describing what we have implemented in plain terms rather than in the abstract.
8.6 We would rather publish an open item than a compliance claim we cannot yet support.
9. Your own obligations when you publish
9.1 We are the provider of the AI system. When you publish its output, you may also carry obligations of your own, as a deployer and as the publisher.
9.2 In particular, where you publish AI-generated or AI-manipulated text on matters of public interest — news, current affairs, politics, health, elections, matters of public debate — without meaningful human review and editorial responsibility, transparency obligations under the EU AI Act may fall on you to disclose that the content is artificially generated.
9.3 Platform rules may also require disclosure of AI-generated content, separately from the law. Those rules are between you and the platform.
9.4 Meaningful human review is the point at which this shifts. If you read, check, edit and take editorial responsibility for a draft, you are doing what the law expects of a publisher. If you copy and post unread, you are not.
9.5 This is not legal advice. If you publish at scale, or on matters of public interest, take your own advice.
10. What we do not claim
We are being deliberate here, because vague reassurance in a disclosure is worse than none.
10.1 We do not claim that the model is accurate.
10.2 We do not claim that the model is unbiased.
10.3 We do not claim that the model is safe, neutral, objective, fair or free from error.
10.4 We make no promise about the quality, originality, legality or fitness for any purpose of anything Ingy generates.
10.5 What we do say is section 4: check it before you publish it, because you are the publisher.
11. Changes to this disclosure
11.1 We may update this disclosure, particularly as the AI Act framework and guidance develop. The version number and effective date at the top will change.
11.2 Material changes to how Ingy generates content, or to which AI providers we use, will be reflected here.
12. Contact
12.1 Questions about this disclosure: [email protected].
12.2 Formal legal notices: [email protected], or IX Labs, 27 Old Gloucester Street, London, WC1N 3AX, United Kingdom.
12.3 This disclosure is governed by the law of England and Wales.
Contact
IX Labs
27 Old Gloucester Street, London, WC1N 3AX, United Kingdom
General enquiries: [email protected]
Legal notices: [email protected]