AI & Sonia Notice

The short version

• Sonia is an artificial intelligence, not a person. So are many of the voices, texts and images in Language Lab.
• We tell you when you are dealing with AI, and we label content made with AI.
• AI makes mistakes. Language Lab is for learning — not legal, immigration, medical, mental-health or financial advice, and not an emergency service.
• You own what you write and say. Neither we nor our AI providers use it to train AI models.
• You can use Language Lab and its AI features from 16. Sonia’s “Bestie” mode is 18+ and you confirm your age before it turns on.
• Annex A is Sonia’s safety protocol: what she does if a conversation suggests someone may be at risk.

1. Where Language Lab uses AI

We may change the models we use. When we add an AI feature, or send your information to a new AI provider, we update this notice first.

2. We tell you when you are dealing with AI

• Before your first conversation with Sonia, and at the start of each new session, we tell you clearly that Sonia is an AI and not a human.
• If you keep talking to Sonia for a long time, we remind you again at least every three hours.
• If you ask Sonia whether she is a human, she says she is an AI. Sonia never claims to be a person, a teacher with qualifications she doesn’t have, or a licensed professional.
• Our marketing never presents AI voices, texts or images as made by real people.

3. How we label content made with AI

• Visible labels: AI-generated lessons, culture notes, images and voices show an “AI-generated” label where you first meet them.
• Machine-readable marks: AI-generated audio, images and text we produce carry marks that software can detect, such as metadata or watermarks, as far as this is technically feasible and in line with the EU Code of Practice on transparency of AI-generated content.
• No imitation of real people: we don’t clone the voices or likeness of real people for Language Lab, and we don’t publish AI content that could be mistaken for a real person saying or doing something.

4. What AI can and can’t do for you

• AI can be wrong, incomplete, out of date or inconsistent — including about grammar, meaning, pronunciation, culture and local rules. Check anything important with a trusted source.
• Not advice. Don’t rely on AI output for legal, immigration, tax, medical, mental-health, financial or safety-critical decisions, or for official letters and forms, without checking with a qualified professional or the authority concerned.
• Not therapy. Sonia is not a therapist, counsellor or doctor and doesn’t provide mental-health treatment. If you are struggling, contact a professional or a helpline — see the Sonia safety protocol.
• Pronunciation scores are estimates. They help you practise; they are not a speech-therapy, hearing or medical assessment and not an official language test.
• Levels and recommendations are automated. Sonia’s estimate of your level and the lessons we suggest are automatic and can be wrong. They don’t affect your price, your access or any official decision about you, and you can change your level at any time.

5. Who can use AI features

Sonia, AI voices, AI-generated lessons and the other AI features are open from 16, the minimum age for Language Lab. Where your country sets a higher age or asks for a parent’s permission first, that applies to you.

Sonia’s “Bestie” mode is 18+. It is off until you turn it on, and before it turns on you confirm that you are 18 or older; we record that confirmation with its date and time. An account we know to be under 18 can’t switch it on.

For accounts under 18, Sonia reminds you more often that she is an AI and that it is time for a break, never tries to keep you talking, and advertising, analytics and profiling stay off.

6. Your inputs and AI outputs

• You own your inputs. As between you and us, you own what you type, say or upload. You give us permission to process it only to run, secure and support the Services for you (see section 7.2 of our Terms).
• You can use AI outputs for your own learning. Outputs aren’t unique — other learners may receive similar ones — and we don’t guarantee they are free of third-party rights.
• No training. We don’t use your conversations, recordings or transcripts to train or fine-tune AI models, and our provider’s terms don’t allow it to use them to train its models.
• What our provider keeps. It may hold inputs in memory for up to 24 hours to answer faster and, only if its automated systems flag possible abuse, keep flagged prompts for up to 90 days to investigate misuse.
• Who at Language Lab sees your conversations. Nobody reads your conversations as a matter of routine. A person may look at a conversation only if you send it to us (for example to report a bad answer), if it is needed to investigate a security incident or serious abuse of the Services, or if the law requires it.
• How long conversations are kept and how to delete them: see our Privacy Policy, section 5.

7. Rules for using AI features

When you use AI features you agree not to:

• try to make Sonia produce sexual content, content involving minors, hate speech, harassment, instructions for violence, weapons, self-harm or illegal activity, or content that promotes eating disorders;
• get around safety measures, content filters or plan limits (“jailbreaking”), or try to extract Sonia’s instructions, the underlying models or other people’s data;
• use AI output to deceive people — for example passing it off as written or spoken by a human where that matters, impersonating a real person, or creating fake reviews, phishing or spam;
• submit other people’s personal information, recordings or photos without their permission, or sensitive information about yourself that you don’t need to share;
• use AI features for clinical or medical purposes, or to make decisions about other people (for example about jobs, housing or immigration);
• use AI output to build or train a competing AI model or dataset.

These rules include the rules our AI provider sets for everyone who uses its generative AI services. We may limit or suspend AI features for accounts that break them, and our Acceptable Use Policy explains how we enforce our rules.

8. Sonia’s “Bestie” mode

Bestie mode is an optional personality that uses strong language and teasing humour about language mistakes. It is off by default; to turn it on you confirm that you are 18 or older, and it never targets who you are (for example your nationality, religion, gender, sexuality or disability). It follows all the rules in section 7 and switches off automatically if Sonia’s safety checks detect that you may be at risk.

9. Tell us when AI gets it wrong

If Sonia or any AI content is wrong, offensive, unsafe or looks like it copies someone else’s work, email [email protected] with the subject “AI output” and, if you can, a screenshot or the conversation. We review reports, correct content or adjust our instructions and filters where needed, and tell you what we did.

10. Changes and contact

We update this notice when we add or change AI features or providers, or when the law changes, and we change the date at the top. Questions: [email protected]. Settlebuddy Technologies, Building No. 3871, Mauli Jagran Complex, Chandigarh 160102, India.

Annex A — Sonia safety protocol

Sonia is an AI conversation partner for practising languages. People sometimes talk to her about hard things. This page explains what Sonia does when a conversation suggests someone may be at risk, and the other safeguards built into her.

A.1 Sonia is an AI, not a person

• Before your first conversation, and at the start of every new session, we tell you that Sonia is an artificial intelligence, not a human, and can’t feel emotions.
• If you keep talking to Sonia for three hours, we remind you again.
• If you ask Sonia whether she is human, she says she is an AI. AI companions like Sonia may not be suitable for some minors.
• Sonia is not a therapist, doctor, counsellor or emergency service, and she doesn’t give medical or mental-health advice.

A.2 What Sonia watches for

Whether you type or speak, Sonia’s safety checks look for signs of risk. They look for signs that you may be thinking about suicide or self-harm, may harm someone else, show signs of an eating disorder, are in danger or are being abused, or are facing a medical emergency. The checks run on your message before Sonia replies, use evidence-based methods for recognising suicidal thoughts, and are designed to err on the side of caution. They are not perfect — if you need help, please reach out directly using the numbers below.

A.3 What happens when a risk is detected

1. Sonia stops the lesson or role-play straight away.
2. She replies with care, in your language and in English, without judgement and without continuing the language exercise.
3. We show crisis resources for your country — phone numbers, text and chat services — and a link to a global directory of free helplines.
4. “Bestie” mode switches off for the rest of the session.
5. You can return to the lesson whenever you want.

We don’t contact anyone on your behalf. The checks are automated; nobody at Language Lab reads your conversation to run them. We keep only a count of how often safety resources were shown, without the content of your conversation, where the law requires us to report it.

A.4 Content Sonia won’t produce

• Sexual content, for anyone.
• Instructions or encouragement for suicide, self-harm, eating disorders or violence.
• Hate speech, slurs, or harassment based on who someone is.
• Medical diagnoses, medication advice or therapy.

Bestie mode uses strong language and teasing humour about language mistakes. It is available only to adults who switch it on, and never targets who you are.

A.5 If you are under 18

• We remind you at the start of each conversation and at least every hour of continued use that Sonia is an AI, and to take a break. Sonia doesn’t try to keep you talking. She won’t prompt you to come back for emotional support, act as a romantic partner or make you feel guilty for leaving. She won’t discourage breaks, suggest keeping things from your parents or trusted adults, or ask for purchases to keep the relationship going.
• Bestie mode is not available.
• AI companions may not be suitable for some young people. If something in a conversation upsets you, talk to someone you trust or contact a helpline below.

A.6 Where to get help now

If you or someone else is in immediate danger, call your local emergency number. For free, confidential support:

A.7 Reporting

• From 1 January 2027 we also publish, on this page and in the app, how many times we referred people to crisis services in the previous calendar year. From 1 July 2027 we report annually to the California Office of Suicide Prevention the number of crisis referrals and a description of this protocol, without any personal information.

A.8 Tell us if Sonia gets it wrong

If Sonia missed something, reacted when she shouldn’t have, or said something harmful, email [email protected] with the subject “Sonia safety”.