manufact / MoteOS Disclaimer and Terms of Use

Version 2026-08-25Effective date [TBC: launch date. For reference, the corporate website privacy policy took effect on 2026-08-07]

⚠️ This is a DRAFT. It must not be published as binding terms, or used to obtain user consent, until it has been reviewed by counsel. Items marked [TBC: …] are open placeholders and must all be resolved before launch. A consolidated list appears in Section 18.

1. Read this part first

manufact is not a chat box that types back at you. It is a system that acts on your behalf: AI agents inside a sandbox dedicated to you execute code, create and delete files, reach out to the internet, and call third-party services on their own initiative. That produces efficiency ordinary software cannot, and it carries risks ordinary software does not.

The purpose of this document is to state those risks clearly, not to bury them under boilerplate. Please read all of it, and Sections 3, 4, and 5 in particular.

2. Scope and acceptance

These Disclaimer and Terms of Use (the “Terms”) govern your use of the manufact service, the MoteOS desktop and browser clients, and our related websites.

“We” and “us” in these Terms means Archaea AI, Inc. (149 Commonwealth Dr, Ste 1090, Menlo Park, CA 94025, USA).

By creating an account, signing in, or using the service, you confirm that you have read, understood, and accept these Terms. If you do not accept them, do not use the service.

Age requirement: you must meet the minimum age where you are — 13 in the United States (consistent with COPPA), the local digital age of consent in the European Economic Area and the United Kingdom (member states set it between 13 and 16), and the minimum age set by local law elsewhere. See Section 13 of the Privacy Policy.

Read these Terms alongside the Privacy Policy. If you use the service on behalf of an organisation, you represent that you are authorised to do so and that the organisation is equally bound.

[TBC: the relationship between these Terms and any future formal Terms of Service or Subscription Agreement, and with the existing Product License page on the corporate website (https://www.archaea-ai.com/product-license), and which prevails on conflict. We recommend consolidating into a single Terms of Service in the final version, or stating the order of precedence explicitly.]

3. About AI output: it can be wrong

3.1 Output may be inaccurate, outdated, or simply false

manufact is built on large language models. Large language models produce content that reads as plausible but is wrong — commonly called hallucination. Output may:

invent facts, citations, statutes, figures, APIs, or functions that do not exist;
be out of date, since a model’s knowledge has a cutoff;
get arithmetic, unit conversions, or date calculations wrong;
misread your intent, or drift away from your original requirement over a long conversation;
state something false with complete confidence — the confidence of the tone tells you nothing about the correctness of the content.

Output is provided as-is. We make no warranty as to its accuracy, completeness, currency, fitness, or reliability.

3.2 It is not professional advice

Output from manufact does not constitute, and must not be treated as, legal, medical, psychological, financial, investment, tax, accounting, insurance, engineering, safety, compliance, or any other form of professional advice.

We are not your lawyer, doctor, accountant, or adviser, and using the service creates no professional relationship. For decisions in those areas, consult a qualified professional.

3.3 You must verify

Before you act on any output in a way that has consequences, it is your responsibility to verify it independently. In particular:

code must be reviewed and tested before it is merged or deployed — AI-generated code can contain logic errors, security vulnerabilities, licence-incompatible fragments, or references to packages that do not exist;
facts, figures, and citations must be traced back to their original sources;
anything touching money, legal obligations, personal safety, or irreversible operations must have a human making the final decision.

We are not liable for losses arising from reliance on unverified output.

3.4 Output is not deterministic

The same input may produce different output. We make no warranty of reproducibility, stability, or consistency across runs.

4. About agent autonomy: it acts on its own

This is the fundamental difference between manufact and an ordinary AI chat product, and it is important that you understand it.

4.1 What an agent will do

Once you give an instruction, an agent may, without asking you to confirm each step:

run code and shell commands;
create, modify, overwrite, or delete files and directories in your sandbox file volume;
install packages and dependencies, including third-party code from public registries;
access the internet: fetch pages, call APIs, download files, run searches;
operate browsers and application interfaces, including filling in and submitting web forms;
use connectors you have authorised, reading and potentially writing your data on third-party services such as code repositories and databases;
run on a schedule you configured, while you are not present.

By giving an instruction, you authorise the agent to take these actions in order to carry it out.

4.2 The risks you accept

Data can be modified or deleted. An agent may misread your intent, delete files you wanted to keep, or overwrite content in a way you did not expect. Some of this is not reversible.
Effects outside the system can be unintended. An agent may send requests to third-party services, submit forms, create resources, or change repository contents. Those actions have already happened in the outside world, and whether they can be undone is up to the other party.
Insecure code can be introduced. Dependencies an agent installs, or code it writes, may contain vulnerabilities or malicious components.
Costs can be incurred. Third-party APIs, cloud resources, or subscription services an agent calls may charge by usage. Those charges are yours.
Third-party terms can be breached. Sites and APIs an agent visits may prohibit automated access. Ensuring your use complies with their terms is your responsibility.

Keep your own backups of anything important, somewhere outside manufact;
Grant the minimum: give an agent only the connectors and permissions the task requires, and revoke them afterwards;
Do not hand production credentials to an agent unless you fully understand and accept the consequences;
Stay present for irreversible operations; do not put destructive actions on a schedule;
Review what an agent produces, especially anything destined for a codebase or for publication.

To the maximum extent permitted by law, we are not liable for data loss, data corruption, unintended external operations, third-party charges, or other consequences of autonomous agent behaviour.

5. About the sandbox: a security boundary, not a compliance boundary

5.1 What isolation we provide

Each user’s agents run inside a dedicated, kernel-level gVisor sandbox container with a private file volume. Sandboxes and file volumes are not shared between users. We consider this a comparatively strong form of isolation for running untrusted code.

5.2 It is not compliance-grade isolation

Sandbox isolation is an engineering security measure. It is not a certification. Specifically:

we hold no HIPAA, PCI DSS, SOC 2, ISO 27001, or comparable certification or audit opinion; [TBC: if any certification has been obtained or is in progress, state its current status and scope accurately here]
we do not act as a HIPAA business associate and do not sign BAAs;
our service is not a PCI DSS compliant environment;
we offer no data residency commitment, no dedicated instances, and no customer-managed encryption keys.

5.3 So do not use it for the following data

Unless we have signed a separate written agreement with you, you must not process the following in manufact:

protected health information (PHI) and other data subject to strict healthcare regulation;
payment card data (full card numbers, CVV, track data);
government classified or restricted information;
personal information of children subject to COPPA or comparable regimes;
biometric identifiers;
any category of data for which applicable law requires specific security controls, certifications, or contractual arrangements that we have not put in place.

If you breach this section, the resulting consequences and liability are yours, and we reserve the right to suspend your account immediately.

5.4 No isolation is absolute

No isolation technology is absolute. Sandbox escapes, supply chain attacks, and zero-day vulnerabilities can all cross the boundary. We patch known vulnerabilities promptly, but we do not warrant that the sandbox cannot be breached.

6. About the pre-release: this is a beta

The service is currently in invite-only pre-release (beta). You understand and agree that:

There is no service level agreement. We make no warranty of availability, uptime, performance, or responsiveness.
It may change or stop at any time. We may modify, suspend, or discontinue any feature without prior notice, including features you rely on.
Access can be revoked. We may withdraw your invitation and access at any time, for abuse, capacity, or any other reason.
Data may be reset or lost. We rebuild environments, change data structures, and migrate infrastructure frequently. Your sandbox, files, and conversation history may be reset or permanently lost. Keep your own backups of anything important.
Features may be unstable. Pre-release builds contain incomplete and insufficiently tested functionality, and may crash, produce inconsistent data, or behave unexpectedly.
It is not suitable for production use. Do not make manufact a dependency of any production system, business-critical process, or sole store of important material.

7. About credits

Credits are an internal metering unit. They are not currency, not a token, not a prepaid instrument, and do not constitute stored value or any kind of financial instrument.
Credits are not redeemable for cash, not transferable, not movable between accounts, and not tradeable.
The rate at which credits are consumed depends on the model used and the complexity of the task. We may change the conversion rules and rate table at any time, and doing so creates no obligation to compensate for credits already consumed.
Credits do not roll over between periods; any unused balance lapses at the end of a plan period.
Credits granted during the beta are a free gift. They are non-refundable, have no cash value, and confer no entitlement.
If your account is suspended or terminated for breach of these Terms, any remaining credits lapse.

[TBC: the refund policy once paid plans launch — cooling-off period, pro-rata refunds, dispute handling, and chargeback response. This section must be consistent with the final billing policy and with the rules of the chosen payment provider (e.g. Paddle or Stripe).]

8. Acceptable use

8.1 Prohibited uses

You must not use manufact — including by instructing an agent — to do any of the following.

Unlawful and infringing conduct

Break any applicable law or regulation, or infringe the rights of others;
Infringe copyright, trade marks, patents, trade secrets, or other intellectual property;
Violate anyone’s privacy, reputation, or likeness rights;
Process personal data you have no right to process.

Attacks on security

Develop, distribute, or deploy malware, ransomware, spyware, or botnets;
Conduct unauthorised penetration testing, vulnerability scanning, brute forcing, or denial-of-service attacks against any system;
Circumvent authentication, access controls, rate limits, or paywalls on any system;
Attack, probe, or attempt to escape manufact’s own sandbox, rate limiting, or quota mechanisms.

Abuse and circumvention

Use automation to scrape third-party sites or APIs at scale, particularly where the operator prohibits it;
Use multiple accounts, false identities, forged or tampered device information (including spoofed MAC addresses, hostnames, or machine ids), or any other means to farm free quota, evade quotas, or evade an account ban — the device identifiers we collect to detect this, and their limits, are set out in Section 6 of the Privacy Policy;
Resell, rent, time-share, or sublicense your account or API access, or repackage the service as a competing product; [TBC: whether building outward-facing products on top of the service is permitted, and on what licensing terms]
Reverse engineer or decompile the service, or attempt to extract model weights, system prompts, or other proprietary material, except where the law expressly permits it.

Harmful content

Generate or distribute child sexual abuse material (CSAM) — we have zero tolerance for this and will report it to the authorities;
Generate non-consensual sexualised content, or deepfakes of real people without their consent;
Generate content for harassment, bullying, hate speech, incitement to violence, or terrorism;
Generate content for fraud, phishing, false identities, academic dishonesty, or election interference;
Provide meaningful assistance in making weapons, explosives, dangerous chemicals, or biological agents.

High-risk applications

Use output, without qualified human review, in medical diagnosis or treatment decisions, legal adjudication, credit or insurance underwriting, employment or admissions decisions, law enforcement or judicial decisions, critical infrastructure control, or any safety-of-life system.

8.2 Consequences of breach

Where we reasonably believe you have breached this section, we may — without prior notice — remove the content, restrict or suspend functionality, suspend or terminate your account immediately, preserve relevant evidence, and report the matter to law enforcement where required or permitted.

We have no obligation to monitor your use, but we reserve the right to do so.

9. Ownership of content

9.1 Your content

You keep all rights in your content. We claim no ownership of your inputs, the files you upload, or the contents of your sandbox.

So that we can deliver the service, you grant us a limited, non-exclusive, royalty-free licence to store, copy, transmit, display, and process your content for the purpose of operating, maintaining, securing, and delivering the service to you — including transmitting it to the third-party providers described in Section 5 of the Privacy Policy. This licence is strictly limited to service delivery and ends when you delete the content or close your account (subject to the ordinary expiry of backups).

We do not rely on this licence to train or fine-tune models, and we do not provide your content to third parties for training — see Section 5.3 of the Privacy Policy. (The one exception is where you use your own model key: whether that upstream provider trains on your content is governed by your agreement with them — see Section 5.2 of the Privacy Policy.)

9.2 AI-generated output

For output an agent generates for you — code, text, images, audio:

As between you and us, we claim no ownership of output, and to the extent permitted by law we assign to you any interest we may hold in it.
But note three important limitations:

(a) It may not be protected by copyright. The current position in several jurisdictions, including the United States, is that purely AI-generated material lacking sufficient human authorship may not attract copyright protection at all. That means you may be unable to stop others from using identical or similar material. The law here is still developing.

(b) Others may receive substantially the same output. Generative models produce similar output for similar input. Other users may receive content that is substantially identical or highly similar to yours. We give no exclusivity warranty and make no promise of uniqueness.

(c) It may resemble existing works. Output may inadvertently resemble an existing protected work, or contain fragments derived from training data. Before using output commercially, it is your responsibility to assess its usability and infringement risk, including running licence compliance checks over generated code.

We give no warranty as to the copyrightability, originality, non-infringement, or commercial fitness of output.

9.3 Our intellectual property

manufact, MoteOS, and the associated software, interfaces, designs, trade marks, and documentation are owned by us or our licensors. These Terms grant you no rights beyond use of the service.

9.4 Feedback

If you send us suggestions or feedback about the product, we may use it freely, with no obligation of payment or confidentiality.

10. Third-party services and open source

10.1 Third-party services

The service depends on and integrates a number of third parties: model providers, speech providers, image and video generation providers, identity providers, an email provider, and infrastructure providers. You may also connect your own services through connectors (repositories, databases, remote MCP servers).

These services are outside our control. We are not responsible for their availability, performance, security, data practices, or changes to their terms.
Outages, changes, or discontinuation of a third-party service may affect functionality, and we are not liable for that.
When you use a third-party service, its own terms apply to you and it is your responsibility to comply with them.

10.2 Marketplaces and user-published content

Some content in the app, theme, blueprint, and MCP marketplaces is published by third parties. We may review it automatically or manually, but:

review is not a warranty of quality, safety, or legal compliance;
installing and using third-party content is at your own risk;
we are not liable for harm caused by third-party content, and we reserve the right to remove any listing at any time.

10.3 Open source components

The service and the clients include open source components under their respective licences. Those components are provided under their own licence terms, generally without warranty of any kind. Licences and notices are available within the client. [TBC: where open source notices are displayed and how to obtain them]

11. Disclaimer of warranties

To the maximum extent permitted by applicable law:

The service and all output are provided “AS IS” and “AS AVAILABLE”, without warranty of any kind, express or implied, including without limitation any implied warranty of merchantability, fitness for a particular purpose, title, non-infringement, accuracy, reliability, security, or uninterrupted operation.

We do not warrant that the service will be uninterrupted, timely, secure, or error-free; that defects will be corrected; that the service or its servers are free of viruses or harmful components; or that output will meet your requirements or any particular standard.

No oral or written information obtained from us or through the service creates any warranty not expressly stated in these Terms.

Some jurisdictions do not allow the exclusion of certain implied warranties, so parts of the above may not apply to you. Where that is the case, any such warranty is limited to the shortest period permitted by law.

12. Limitation of liability

To the maximum extent permitted by applicable law:

(a) Neither we nor our affiliates, directors, employees, agents, or licensors will be liable for any indirect, incidental, special, consequential, or punitive damages, including without limitation lost profits, lost revenue, lost goodwill, loss of or damage to data, business interruption, or the cost of substitute services — even if we have been advised of the possibility.

(b) Our total aggregate liability for all claims arising out of these Terms or the service will not exceed: [TBC: the liability cap. A common formulation is "the greater of (i) the fees you actually paid us in the 12 months preceding the event giving rise to the claim, and (ii) USD 100." During the beta no fees are paid, so the effective cap is the fixed amount — please confirm.]

(c) These limitations apply across all theories of liability (contract, tort, strict liability, or otherwise) and survive any failure of essential purpose of a limited remedy.

(d) These limitations do not apply to liability that cannot be excluded or limited by law, including our fraud or wilful misconduct, and death or personal injury caused by our negligence.

(e) Some jurisdictions do not allow the exclusion or limitation of certain damages, so parts of the above may not apply to you. If you are a consumer in the EU or the UK, your mandatory consumer protection rights are unaffected by these Terms.

13. Indemnity

To the extent permitted by applicable law, you agree to indemnify, defend, and hold harmless us, our affiliates, employees, and agents from third-party claims, losses, liabilities, damages, and reasonable costs (including reasonable legal fees) arising out of:

your use of the service, or actions you instruct an agent to take;
your breach of these Terms, in particular the data restrictions in Section 5.3 and the acceptable use rules in Section 8;
your content, or your use of output, infringing the rights of a third party;
your breach of applicable law or of a third party’s terms.

We reserve the right to assume exclusive defence and control of any matter subject to indemnification, and you agree to cooperate.

[TBC: in most jurisdictions an indemnity of this kind is unenforceable or of limited effect against consumers. Counsel should confirm the scope for a consumer-facing service and consider distinguishing business users from consumers.]

14. Suspension and termination

You may stop using the service and request account closure at any time (see Section 11 of the Privacy Policy for how).
We may suspend or terminate your account immediately for breach of these Terms. In other circumstances we will make reasonable efforts to give notice.
During the beta we may withdraw access for capacity reasons, feature changes, or any other reason.
On termination: your access ends immediately; any remaining credits lapse; your data is handled under the retention rules in Section 10 of the Privacy Policy. Export anything you need to keep before terminating.
Provisions that by their nature should survive termination (Sections 9, 11, 12, 13, and 15 among them) survive.

15. Governing law and disputes

Our legal entity, Archaea AI, Inc., is registered and based in Menlo Park, California, USA. This section therefore sets out a recommendation, but it is not yet settled:

Governing law: we recommend the laws of the State of California, excluding its conflict-of-laws rules. [TBC: counsel to confirm. Note that a choice of law cannot deprive a consumer of the protection of mandatory rules in their country of residence — for consumers in the EU, the UK, Japan, and Korea, local mandatory rules may prevail.]
Dispute resolution and venue: we recommend the courts of California (state or federal), or arbitration seated in California. [TBC: counsel to confirm litigation or arbitration. If arbitration, specify the institution, seat, language, and rules; whether a class action waiver applies (noting it may be unenforceable in the EU and elsewhere); and whether a small-claims carve-out applies.]
Informal resolution first: before starting formal proceedings, both parties will try in good faith to resolve the matter through sales@archaea-ai.com (to be replaced with a dedicated legal address before launch — see Section 18), over a period of [TBC: 30 days suggested].
EU consumers: you may use the EU Online Dispute Resolution platform or approach the consumer protection body in your member state. You always retain the right to bring proceedings in the courts of your place of residence.

16. General

Severability: if any provision is held invalid or unenforceable, it will be modified or severed to the minimum extent necessary, and the rest remains in force.
No waiver: our failure or delay in exercising a right is not a waiver of it.
Assignment: you may not assign these Terms without our written consent; we may assign them in a merger, acquisition, or sale of assets.
Entire agreement: these Terms and the Privacy Policy are the entire agreement between you and us regarding the service, superseding all prior communications.
Force majeure: we are not liable for delay or failure to perform caused by events beyond our reasonable control (natural disasters, war, strikes, network or power outages, government action, third-party provider failures).
Language: these Terms exist in English and Chinese. We recommend the English version prevails, consistent with the California governing law recommended in Section 15. [TBC: counsel to confirm. Note that a consumer-facing service in the EU may need to provide, and be bound by, the consumer's own language.]

17. Changes and contact

We may revise these Terms from time to time. Material changes will be notified at least 30 days in advance by email or a prominent in-client notice; continuing to use the service after a change takes effect constitutes acceptance. If you do not accept, stop using the service and close your account.
The version is shown in the version field at the top of this document. We record which version you accepted.
Contact: sales@archaea-ai.com; postal address: Archaea AI, Inc., 149 Commonwealth Dr, Ste 1090, Menlo Park, CA 94025, USA
Abuse reports: the same address. (It is the only address we currently publish; setting up dedicated legal and abuse aliases before launch is an operational task recorded in Section 18.)

18. Open placeholders for the business and counsel

Consolidated list of every [TBC] above. This document must not go live until all of them are resolved.

Three items settled on 2026-08-25 and no longer open: (1) the legal entity is Archaea AI, Inc. — its name, address, and contact details are filled in at §2 and §17 per the corporate website; (2) content is not used for training (§9.1, BYOK excepted, matching Privacy Policy §5.3); (3) device identifiers are stored in clear text — the “forged device information” prohibition in §8.1 pairs with Section 6 of the Privacy Policy.

Effect and document relationships

Effective date (frontmatter effective_date; the corporate website policy’s effective date of 2026-08-07 is available for reference)
Relationship and order of precedence between these Terms, any future formal Terms of Service or Subscription Agreement, and the existing Product License page on the corporate website (§2)
Which language version prevails on discrepancy — English recommended, counsel to confirm (§16)

Jurisdiction and disputes (narrowed to “recommendation pending confirmation”; counsel to finalise) 4. Governing law: California recommended (the entity is based in Menlo Park, CA); confirm how consumers’ local mandatory rules interact (§15) 5. Dispute resolution and venue: California courts or California-seated arbitration recommended; confirm litigation vs arbitration, the institution and rules, class action waiver, and small-claims carve-out (§15) 6. Length of the informal resolution period (§15) 7. Scope of the indemnity for consumers, and whether to distinguish business users from consumers (§13)

Liability and commercial terms 8. The specific liability cap (§12) 9. The refund policy once paid plans launch — cooling-off, pro-rata refunds, disputes and chargebacks; must match the payment provider’s rules (§7) 10. Whether building outward-facing products on the service is permitted, and on what licensing terms (§8.1)

Compliance and disclosure 11. Whether any security or compliance certification has been obtained or is in progress (SOC 2, ISO 27001, and so on); if so, state its status and scope accurately (§5.2) 12. Where open source notices are displayed and how to obtain them (§10.3)

Operational tasks before launch (not marked as placeholders in the body, but required before going live)

Set up dedicated email aliases. The only address published today is sales@archaea-ai.com, which is what the body of this document uses. Before launch, set up legal@archaea-ai.com (legal, terms, and informal resolution) and abuse@archaea-ai.com (abuse reports), route them to a ticketing system, and replace the address in §15 and §17 here and in the Privacy Policy (privacy requests get a separate privacy@). Until those aliases actually work, they must not appear in the body text.
Implement the age threshold in the product. The age requirement in §2 has to match the sign-up flow — whether an age declaration or verification step is added is tracked as the corresponding placeholder in Section 19 of the Privacy Policy.