Carae — Terms and Conditions

Status: Approved — awaiting solicitor review at a later date.

Effective: TBD on publication Version: tos-v1

⚠️ Pre-launch checklist (see also the "carae-legal-pre-launch-checklist" cron):

1. Replace [Legal entity name TBD] with the registered Irish company name.
2. Replace [Address TBD] with the registered office address.
3. Replace "TBD on publication" with the actual go-live date.
4. Have a solicitor review (Mason Hayes & Curran or Matheson recommended). Operator: [Legal entity name TBD — Durg to register an Irish private limited company before launch] Registered office: [Address TBD] Contact: [email protected]

1. The Service

Carae ("the Service", "we", "our") is a personal AI assistant that operates over messaging platforms (currently Telegram, with WhatsApp and iMessage planned). The Service helps users manage calendars, email, reminders, and other personal-productivity tasks, including by installing and running third-party-connected "Skills" on their behalf.

By creating an account, connecting integrations, or messaging the Service, you ("the User", "you", "your") agree to these Terms.

You must be at least 18 years of age to use the Service. If you are using the Service in the course of employment or for a business, you confirm you are authorised to bind that business to these Terms.

2. Plans, Boosts, and Billing

2.1 Subscription tiers

The Service is offered on the following plans:

Plans, prices, and limits may change with at least 30 days' notice; existing subscribers retain their current price for 90 days from the change date or for the duration of their current annual term, whichever is later.

2.2 Boosts

Some Service operations ("Heavy Actions" — e.g. deep web research, long-form drafting, bulk email summarisation) consume one Boost per use. Boosts reset on the first day of each calendar month for paid plans. Trial Boosts do not reset.

Boosts may also be purchased ad-hoc as top-up bundles. Top-up Boosts do not expire while your subscription is active and are forfeit 90 days after subscription cancellation.

2.3 Daily Skill install caps

Each plan includes a daily limit on the number of new Skills that can be installed. The cap resets at midnight in the User's local timezone. Uninstalling a Skill does not free up a slot in that day's count.

2.4 Billing, refunds, and cancellation

Billing is handled by Stripe. Subscriptions auto-renew unless cancelled before the end of the current period. You may cancel at any time via /plan in the bot or by emailing [email protected]; cancellation takes effect at the end of the current billing period.

We offer a 14-day refund on the first month of any new paid subscription if you have used fewer than 5 Boosts in that period. Beyond that, payments are non-refundable except where required by Irish or EU consumer law.

If a payment fails, your account moves to a 7-day grace period during which the Service continues at reduced capacity. After 7 days, the account is suspended; data is preserved for 30 days before deletion.

3. Skills and Third-Party Integrations

3.1 What Skills do

Skills are pre-built capabilities you can install. They fall into three categories:

• First-party Skills built and maintained by Carae. These are clearly marked in the catalogue.
• Third-party Skills built by external providers (vendors with whom we have a partnership) or community contributors. These are marked with the contributor's name in the catalogue. Third-party Skills go through a verification review before being made publicly available, but we do not warrant their behaviour and the third party (or the contributor) is the data controller for any data their Skill processes outside our infrastructure.
• Custom Skills that you (Pro plan) or someone you trust create. You are responsible for the behaviour of your own Custom Skills.

Some Skills are local to the Service (memory, journal, reminders); some require connecting third-party accounts (Google Calendar, Gmail, etc.). When you install a Skill, you authorise the Service to perform the actions described for that Skill on your behalf.

For third-party and Custom Skills, our liability is limited to ensuring the Skill cannot escape the Service's sandbox or access data belonging to other users. We are not responsible for bugs, errors, or harmful outputs from a third-party or community-contributed Skill beyond removing it from the catalogue once we become aware of an issue.

3.2 General AI-generated content

The Service is built around large language models (Anthropic Claude, OpenAI GPT, Google Gemini, and others). You acknowledge and accept that:

• AI outputs are probabilistic, not authoritative. Every reply, summary, draft, recommendation, or generated piece of content the Service produces is an LLM output. It may be factually wrong, internally inconsistent, biased, or fabricated even when delivered with apparent confidence. You should treat all Service output as a starting point requiring your own verification, not as professional advice or established fact.
• The Service is not a substitute for professional advice. Carae is not a doctor, lawyer, accountant, financial adviser, therapist, pharmacist, or licensed professional in any field. Do not rely on the Service for medical, legal, tax, financial, mental-health, safety, or other professional decisions. If you have a serious health, legal, or financial question, consult a qualified human professional.
• The Service may hallucinate facts. It may invent names, dates, figures, citations, addresses, prices, statutes, case law, or any other detail that sounds correct but is not. We mitigate this with retrieval, web grounding, and routing to specialised models, but no LLM-based system has eliminated hallucination. If accuracy matters, you must verify independently.
• The Service may produce unsafe or upsetting content. Despite our safety measures, an LLM may generate output that is offensive, distressing, or otherwise inappropriate. Use of the Service is at your discretion; if a generation upsets you, stop the conversation and contact us.
• The Service may misunderstand instructions. A casual phrase you intended one way may be interpreted differently by the model. Where the Service is about to take a consequential action (modify your calendar, archive emails, send a message, spend a Boost on a heavy operation), it should ask for confirmation, but you should not assume that a confirmation will always be requested.
• You are responsible for what the Service does on your behalf. When you instruct the Service to perform an action — whether through a Skill, a connected integration, or a conversational message — you are responsible for the consequences of that action, even where the Service's interpretation of your instruction differed from what you intended.

3.3 Skill-specific limitations

In addition to §3.2, you acknowledge and accept that:

• Skills act on your behalf. When you install a Skill that has write access to a third-party account (for example, archiving emails, declining meetings, posting to a calendar), the Service may perform those actions automatically based on Skill rules and your conversational instructions.
• Connected third-party services have their own terms. Google, Microsoft, and other providers may rate-limit, suspend, or revoke access independently of us. The Service's availability of integrations depends on those providers' continued cooperation and may change without notice.
• You are responsible for backups. The Service is not a backup service. Where a Skill modifies or deletes data in a third-party account (e.g. an email auto-archive rule), the responsibility for maintaining backups of that data rests with you. We strongly recommend you have independent backups of any account you connect.

3.4 Limitation on damages

To the maximum extent permitted by Irish and EU law:

The Service, its operators, employees, and contractors will not be liable for any direct, indirect, incidental, consequential, special, or exemplary damages arising from:

• AI-generated content (including but not limited to: summaries, drafts, recommendations, factual statements, code, translations, images, voice transcripts, or anything else produced by an LLM through the Service);
• a Skill performing an action you instructed it to perform (whether the instruction was given clearly, ambiguously, or via an AI misinterpretation that any reasonable user could have caught on review);
• the deletion, modification, archiving, sending, or sharing of data in any third-party account connected to the Service;
• a third-party API failing, rate-limiting, charging you, suspending your account, or being temporarily or permanently unavailable;
• a Skill triggered by message-pattern matching, scheduled cron, or webhook firing at an inopportune moment;
• you connecting an account that you do not have authority to connect (e.g. an employer's account);
• decisions you took, or did not take, in reliance on the Service's output, summary, recommendation, or generated content;
• a third-party or community-contributed Skill misbehaving, beyond our obligation to remove it from the catalogue when we are notified.

This section applies whether the harm arose from a Skill, from general LLM interaction, from an integration, or from any other use of the Service.

3.5 Cap on aggregate liability

Where Irish or EU law would otherwise impose liability on us, our aggregate liability to you for all claims arising in any 12-month period is capped at the greater of (a) the fees you paid to us in the previous 12 months, or (b) €100. This cap is the standard for consumer SaaS in Ireland and the EU; without it, our insurance and the viability of offering the Service at consumer prices would not be possible. This cap does not apply to liability for death or personal injury caused by negligence, fraud, fraudulent misrepresentation, or any other liability that cannot lawfully be limited.

4. Acceptable Use

You agree not to use the Service to:

• send, generate, or store content that is illegal under Irish or EU law;
• send unsolicited bulk messages, spam, or commercial communications to third parties;
• circumvent rate limits, install caps, or boost-consumption controls;
• impersonate another person, including by connecting accounts you do not own;
• attempt to extract, copy, or reverse-engineer the Service's prompts, internal models, or skills system;
• use the Service to scrape, harvest, or attack any third-party service in a way that violates that service's terms;
• automate the Service for use on behalf of more than one human (one user, one account);
• use the Service in any safety-critical context (medical decision-making, emergency response, financial trading, vehicle control).

We may suspend or terminate accounts that violate this section without notice or refund.

5. Account Termination

5.1 By you

You may delete your account at any time using /forgetme in the bot or by visiting [carae.ai/account/delete]. Deletion is permanent and triggers cascade-removal of your messages, memories, integrations, scheduled jobs, and audit log entries within 24 hours of the request, in accordance with Section 6 of our Privacy Policy. A short-lived deletions_log entry is retained (with no PII) for fraud-prevention and audit purposes.

5.2 By us

We may suspend or terminate your account for violation of these Terms, for non-payment, or where required by law. Where the suspension is for non-payment, we will provide a 7-day grace period before suspension and 30 days before deletion. Where the suspension is for violation, we may terminate immediately if the violation is severe (e.g. illegal content, attempted account compromise).

6. Privacy

Use of the Service is governed by our Privacy Policy, which forms part of these Terms by reference. By accepting these Terms you also accept the Privacy Policy.

7. Changes to these Terms

We may update these Terms. For changes that materially reduce your rights or increase your obligations, we will notify you in-app at least 30 days before the change takes effect. By continuing to use the Service after the effective date, you accept the updated Terms. If you do not accept, you may cancel and request deletion of your account before the effective date.

The version of these Terms in force at any time is identified by the tos-vN tag; we record the version you accepted at signup so we can show you which Terms apply to you.

8. Community Skills (The Forge)

Carae includes The Forge, a community skill marketplace where users can build and share skills with other Carae users.

• Community skills are created by Carae users, not by Carae Ltd. They are not endorsed, warranted, or officially supported by Carae.
• Carae reviews all community skills for safety before publication, but cannot guarantee their behaviour. Community skills are clearly labelled and distinct from Official Carae skills.
• By installing a community skill, you acknowledge that its behaviour may differ from official Carae skills. You install community skills at your own risk.
• Carae reserves the right to remove, suspend, or modify any community skill at any time, without notice, if it is found to violate these Terms or Carae's Community Guidelines.
• Skill creators are solely responsible for the content, accuracy, and behaviour of their submissions. By publishing a skill to The Forge, you confirm it complies with these Terms and does not infringe on any third-party rights.
• To the fullest extent permitted by applicable law, Carae's liability for any harm arising from community skill behaviour is limited to the amount paid by the affected user for Carae services in the 30 days preceding the incident.

9. Governing Law and Disputes

These Terms are governed by the laws of Ireland. The courts of Dublin, Ireland have exclusive jurisdiction over any dispute, except that as a consumer you retain any right to bring proceedings in your country of habitual residence under EU consumer-protection law.

Before commencing legal proceedings, you agree to attempt to resolve any dispute by emailing [email protected] with the subject line "Dispute". We will respond within 14 days.

10. Contact

For questions about these Terms: [email protected] For privacy and data-protection requests: [email protected]

End of Terms — DRAFT v1.