Version 1.6 — 2026-10-11 (v1.5: 2026-10-10; v1.4: 2026-10-09; v1.3: 2026-09-25; v1.2: 2026-09-25; v1.1: 2026-09-24; v1.0: 2026-08-28). Effective upon publication in the Service. Changes in 1.6: self-hosted, cookie-free Umami analytics on public pages (Section 11). Changes in 1.5: Credential Store, agent credential use and HOS Vault (Sections 2, 3, 5, 6, 8). Changes in 1.4: off-site encrypted backups (Backblaze, USA) disclosed in Sections 6, 7 and 8; Microsoft Clarity analytics on public pages only, with consent (Sections 6 and 11); programme owners' visibility of contractor posts (Section 5). Changes in 1.3: when HOS provides the AI access, content is no longer sent to providers whose terms allow training on inputs (Section 3.4). Changes in 1.2: Mistral AI and AIMLAPI training terms clarified (Sections 3.4 and 6). Changes in 1.1: AI sub-processors updated (Section 6), and managed organisation seats explained (Section 5.6).
HOS is an experimental AI personal assistant. We store your chats with the AI and let you connect your own email, cloud storage, and other accounts using your own credentials. To answer you, the AI has to process your messages, so your content is handled in readable form at the moment of processing — but it is encrypted at rest with a key unique to your account, so a stolen database or backup does not expose it. We operate a "zero standing access" policy: nobody on our team reads your private content by default — the only way your content reaches us is if you send it to us in an in-app Support request. We keep your content until you delete it, and you can export or delete your data, individual sessions, or your whole account at any time. We do not use advertising or tracking cookies. This summary does not override the numbered sections below.
1.1. This Privacy Policy explains how EUMENES.AI, a sole proprietorship registered in Singapore under business registration number 53520480M, with a registered address at 331 New Bridge Road, #04-09 Dorsett Residences, Singapore 088764 (the "Operator", "we", "us", "our"), collects, uses, discloses, and protects personal data in connection with "HOS", our AI personal-assistant web application (progressive web app) provided under the Eumenes brand at eumenes.ai and related subdomains (the "Service").
1.2. For the purposes of the EU General Data Protection Regulation and the UK GDPR (together, "GDPR"), we are the controller of the personal data described in this policy, except where we act as a processor on your behalf for content you route through the Service (see Section 4). For the purposes of the Singapore Personal Data Protection Act 2012 ("PDPA"), we are the organisation responsible for the personal data in our possession or control.
1.3. This policy applies to the beta Service, which is invite-only and directed at individual consumers in the EU/UK, Singapore, and worldwide. It should be read together with our Terms of Service and Acceptable Use Policy.
1.4. Our Data Protection Officer (or privacy contact) can be reached at dpo@eumenes.ai.
We collect and process the following categories of personal data.
2.1. Data you provide to create and use your account (account data): - your email address (used for login via magic link and/or passkey, and for transactional emails); - authentication data such as passkey public-key credentials, login tokens/session cookies, and enrolment/device identifiers; - optional profile information you choose to provide (for example, a display name).
2.2. Content you create or connect (user content): - your chats/messages with the AI assistant and the AI's responses; - sessions, notes, files, and other materials you create or upload; - voice recordings you dictate, and the text transcriptions produced from them; - data accessed from, or exchanged with, third-party accounts and services you connect using your own credentials — including your email (via IMAP), cloud storage, and other accounts ("Connected Services"). This may include the contents of your messages, files, and metadata held by those services.
2.3. Voice-clip diagnostic buffer: to diagnose speech-to-text problems, the Service may keep a short, transient buffer of the most recent voice clips and related diagnostic information. This buffer is limited in size and retained only transiently (see Section 8).
2.4. Support content (consent-based): any information you choose to submit to us through an in-app Support request, which may include excerpts of your content that you deliberately attach so we can help you (see Section 5 on zero standing access).
2.5. Technical, telemetry, and diagnostic data: limited operational data generated when you use the Service, such as event/telemetry data about interactions and feature use, diagnostic and error logs, device and browser information, approximate technical information derived from your connection, timestamps, and server logs. We use this to operate, secure, debug, and improve the Service.
2.6. Cookies: only strictly-necessary authentication/session cookies. We do not use advertising or third-party tracking cookies (see Section 11).
2.7. Special-category / sensitive data: the Service is a general-purpose assistant and is not intended for you to submit special categories of personal data (such as health, biometric, political, religious, or sexual-orientation data) or other highly sensitive information. If you choose to include such data in your content, you do so on your own initiative and are responsible for ensuring you have a lawful basis to do so; where GDPR applies, our processing of any such data you submit relies on your explicit consent, which you may withdraw at any time by deleting the content.
Credentials and vault data. Items you save in the Credential Store (site, username, encrypted secret, optional 2FA seed, sharing settings, which agent saved it) and an audit trail of each agent read or write (item label, session, time — never the secret value). For HOS Vault: your account email, encrypted vault data (we cannot read it), device and login metadata (IP address, device type, times), and two-step-login settings.
We process personal data for the purposes below. Where GDPR applies, we rely on the lawful bases indicated.
| Purpose | Personal data | Lawful basis (GDPR) |
|---|---|---|
| Create and administer your account; authenticate you (magic link/passkey); provide the core Service | Account data, authentication data | Performance of a contract (Art. 6(1)(b)) |
| Store your chats, sessions, and files; generate AI responses; provide dictation (speech-to-text) and text-to-speech | User content, voice recordings and transcriptions | Performance of a contract (Art. 6(1)(b)); consent for voice processing (Art. 6(1)(a)) |
| Connect and use the third-party accounts you choose to connect | Connected-Services data, the credentials you supply | Performance of a contract (Art. 6(1)(b)); your consent to connect (Art. 6(1)(a)) |
| Store your credentials and let agents you authorise use them; operate HOS Vault; security auditing of credential access | Credentials and vault data, agent-access audit records | Performance of a contract (Art. 6(1)(b)); security audit: legitimate interests (Art. 6(1)(f)) |
| Diagnose speech-to-text issues (short voice-clip buffer) | Voice-clip diagnostic buffer | Legitimate interests — service reliability (Art. 6(1)(f)); consent to use the microphone (Art. 6(1)(a))|
| Respond to and resolve your Support requests | Support content you submit | Consent (Art. 6(1)(a)); performance of a contract (Art. 6(1)(b)) |
| Operate, secure, monitor, debug, prevent abuse, and improve the Service | Telemetry, diagnostics, logs, technical data | Legitimate interests — security, integrity, and improvement of the Service (Art. 6(1)(f)) |
| Send transactional/service emails (e.g. login links, essential notices) | Email address | Performance of a contract (Art. 6(1)(b)); legitimate interests (Art. 6(1)(f)) |
| Comply with legal obligations and respond to lawful requests | As relevant | Legal obligation (Art. 6(1)(c)); legitimate interests (Art. 6(1)(f)) |
| Establish, exercise, or defend legal claims; enforce our Terms and AUP | As relevant | Legitimate interests (Art. 6(1)(f)); legal obligation (Art. 6(1)(c)) |
3.1. Legitimate-interests balancing. Where we rely on legitimate interests, we have assessed that those interests are not overridden by your interests or fundamental rights and freedoms. You may object to such processing (see Section 12).
3.2. PDPA. Where the PDPA applies, we collect, use, and disclose personal data for purposes that a reasonable person would consider appropriate in the circumstances, and, where required, on the basis of your consent (including deemed consent) or an applicable exception under the PDPA.
3.3. No advertising or profiling for marketing. We do not sell personal data, and we do not use your content or personal data to build advertising profiles or for cross-context behavioural advertising. We do not carry out automated decision-making that produces legal or similarly significant effects on you.
3.4. Model training. We do not use your private User Content to train our own AI models. Your content is processed by the third-party AI providers listed in Section 6 to generate responses, and their practices differ:
When HOS provides the AI access, your content is never sent to a provider whose terms allow it to train on your inputs. If you connect your own key for such a provider, that provider's terms apply to what you send it.
4.1. We are the controller for account data, authentication data, telemetry/diagnostics, logs, and Support content.
4.2. For User Content that you route through the Service and for data exchanged with your Connected Services at your direction, we act largely as a processor / intermediary handling that content on your behalf and under your instructions, using the sub-processors listed in Section 6. You are responsible for having a lawful basis to process any third-party personal data you introduce into the Service.
5.1. Encryption at rest with per-user keys. Your stored content is encrypted at rest using a key unique to your account. This protects your content against theft or compromise of our database or backups: a stolen datastore does not, by itself, expose your content.
5.2. This is at-rest protection, not end-to-end encryption. To generate AI responses, transcribe your voice, and deliver the Service's features, the Service must process your content in plaintext (readable form) at runtime — that is, at the moment it is being processed. We do not claim that your content is end-to-end encrypted or that we are technically incapable of accessing it during processing. We describe our safeguards honestly so you can make informed decisions.
5.3. Zero standing access. As a matter of policy and practice, no member of our team reads users' private content by default. We do not routinely browse, monitor, or read your chats, files, or connected data. The only way your private content reaches a human on our side is if you explicitly submit it to us through an in-app Support request — that is, an affirmative, consent-based act by you. This is a genuine commitment, not merely a statement of intent.
5.4. Limited exceptions. We may access specific content without a Support request only where strictly necessary and lawful: (a) to comply with a valid legal obligation or lawful order; (b) to detect, prevent, or address security incidents, fraud, abuse, or violations of our Terms/AUP (including CSAM and other unlawful content, where detection or reporting is legally required); or (c) to protect the rights, safety, or property of users, the public, or us. Any such access is limited to what is necessary and is handled with confidentiality.
5.5. Organisational and technical measures. We apply access controls, the principle of least privilege, encryption in transit (HTTPS/TLS), tenant isolation between users, authentication safeguards, and logging appropriate to a beta service of this nature. No method of transmission or storage is completely secure, and we cannot guarantee absolute security.
5.6. Managed organisation seats. Sections 5.3 and 5.4 describe private accounts. An organisation may issue a managed seat: an account provided to someone who works for that organisation (for example, a contractor), solely for doing that work. The issuing organisation's owners and administrators have full, standing access to everything created or done in a managed seat, in the same way an employer can access a work laptop it provides. This is shown to the seat holder and must be accepted before the seat can be used. Such access is limited to the issuing organisation; it never extends to our staff or to anyone else, and it never covers private accounts. Do not use a managed seat for personal matters.
5.7. Programme workstreams. If you work as a contractor on someone else's programme in HOS, the programme owner can see everything you post in the workstreams you are assigned to.
Credential Store secrets are encrypted with a per-account key wrapped by a key held outside the application database. They are decrypted only to deliver a value to an agent session you shared it with, or to show it to you. Values are never written to logs or chat transcripts, and the Service redacts any value it delivered if it appears in the chat view. HOS Vault content is end-to-end encrypted on your device; we hold only ciphertext and cannot decrypt it.
6.1. To deliver the Service, we use the following categories of third-party sub-processors. They process personal data only to provide their service to us and are bound by their own terms and, where applicable, data-processing agreements.
| Sub-processor / category | Role | Data processed |
|---|---|---|
| Anthropic, NVIDIA, Groq, OpenAI, Mistral AI SAS (France; not currently in active use), Google LLC (Gemini API) and AIMLAPI (a model aggregator routing to third-party-hosted models; hosting locations not published) — large-language-model / generative-AI providers that generate responses. See Section 3.4 for each provider's training terms. | Generate AI responses, summaries, and other text outputs | The prompts, chats, and content you submit for processing |
| OpenAI | Text-to-speech (spoken playback of text) | The text you choose to have read aloud |
| Groq | Speech-to-text (voice dictation transcription) | Voice recordings/clips you dictate |
| Brevo | Transactional email delivery (magic-link login and essential service emails) | Your email address and the message content of transactional emails |
| Cloud / hosting and infrastructure provider(s) | Host and run the Service and store data (in Singapore) | All data processed by the Service, as hosted |
| Backblaze, Inc. (USA) | Encrypted off-site backups for disaster recovery | Encrypted copies of the Service's data; encrypted before upload with a key Backblaze does not hold, so Backblaze cannot read them |
| Microsoft Corporation (Microsoft Clarity, USA) | Website analytics on public pages, with consent | Analytics/session data from public (signed-out) pages only, via cookies _clck/_clsk |
| Services you connect (Connected Services) — e.g. your email/IMAP provider, cloud-storage provider, and any other account you connect | Provide the connected feature you request | The data you access, send, or exchange with that service at your direction |
6.2. The Connected Services you choose to connect are controlled by their providers under their own terms and privacy policies. We are not responsible for their processing. You should review their policies before connecting.
6.3. We may add, replace, or remove sub-processors as the Service evolves. We will keep this list up to date. Where required, we put in place data-processing agreements and appropriate transfer safeguards with our sub-processors.
6.4. We may also disclose personal data: (a) to professional advisers (legal, accounting) under confidentiality; (b) to authorities, regulators, or courts where required by law or to protect rights and safety; and (c) to a successor entity in connection with a merger, acquisition, or sale of assets, subject to this policy.
7.1. The Service and its data are primarily processed and stored on infrastructure located in Singapore. Some sub-processors (for example, AI providers and email delivery) may process data in other countries. Encrypted disaster-recovery backups are stored with Backblaze in the United States; they are encrypted before they leave our servers. A second encrypted backup copy is kept on operator-controlled hardware in Singapore.
7.2. Where we transfer personal data from the European Economic Area or the United Kingdom to a country not recognised as providing an adequate level of protection, we rely on appropriate safeguards, including the European Commission's Standard Contractual Clauses (and the UK International Data Transfer Agreement / UK Addendum as applicable), together with supplementary measures where needed.
7.3. Where the PDPA applies, we take reasonable steps to ensure that recipients of transferred personal data provide a standard of protection comparable to the PDPA.
7.4. You may request further information about the transfer mechanisms we use by contacting dpo@eumenes.ai.
8.1. User content (chats, sessions, files): retained until you delete it. There is no fixed expiry; your per-user storage quota acts as the natural cap on how much is stored. When you delete content, a session, or your account, we delete the corresponding content from active systems and it is removed from encrypted backups within 30 days, as backups rotate.
8.2. Operational data: retained only transiently or for limited periods appropriate to its purpose, and then deleted or aggregated/anonymised: - login tokens / session cookies — for the life of the session or the cookie's validity period, then expire; - voice-clip diagnostic buffer — only a short, capped buffer of the most recent clips, retained transiently for diagnostics and then overwritten/deleted; - telemetry / diagnostics and server logs — kept for a limited period for security, debugging, and abuse-prevention, then deleted or aggregated.
8.3. Support content: retained for as long as needed to resolve your request and for a limited period afterward for record-keeping, then deleted.
8.4. We may retain limited data longer where required to comply with legal obligations, resolve disputes, prevent abuse or fraud, or establish, exercise, or defend legal claims.
Credential Store items: until you delete them or your account; agent-access audit records: 12 months. HOS Vault: until you delete your vault account or your HOS account; deleted vault data leaves our encrypted backups within the backup retention window (currently up to 15 days).
9.1. Within the Service you can, at any time: export your data; delete individual sessions; delete your content; and delete your entire account. Deleting your account removes your account data and User Content from active systems, subject to the limited retention described in Section 8.4 and the backup-rotation cycle.
9.2. You can disconnect any Connected Service at any time, which stops further access by the Service to that account. This does not delete data already stored by that third party.
HOS Vault can be exported at any time from any Bitwarden client (encrypted JSON recommended); Credential Store items are included in your account export.
10.1. Subject to applicable law and to verification of your identity, you have the following rights.
Under the GDPR (EU/UK), you have the right to: - access the personal data we hold about you and receive a copy; - rectify inaccurate or incomplete personal data; - erase your personal data ("right to be forgotten"); - restrict processing in certain circumstances; - data portability — receive certain data in a structured, commonly used, machine-readable format and, where feasible, have it transmitted to another controller (the in-app export supports this); - object to processing based on legitimate interests, and to object at any time to processing for direct marketing (we do not carry out direct marketing); - withdraw consent at any time where we rely on consent (this does not affect processing already carried out); - not be subject to solely automated decisions with legal or similarly significant effects (we do not carry out such decision-making); and - lodge a complaint with a supervisory authority (see Section 10.4).
Under the PDPA (Singapore), you have the right to: - request access to, and information about the use and disclosure of, personal data about you; - request correction of personal data; and - withdraw consent to the collection, use, or disclosure of personal data, on reasonable notice (withdrawing consent may prevent us from providing the Service).
10.2. How to exercise your rights. Many rights can be exercised directly using the in-app export and deletion controls. Otherwise, contact dpo@eumenes.ai. We will respond within the timeframes required by applicable law (generally within one month under the GDPR and within the statutory timelines under the PDPA), and we may ask for information to verify your identity. Our services are provided free of charge for reasonable requests; we may charge a reasonable fee or decline where a request is manifestly unfounded, excessive, or repetitive, to the extent permitted by law.
10.3. Exercising these rights may, in some cases, limit or prevent our ability to provide the Service.
10.4. Complaints. If you have concerns, we encourage you to contact us first at dpo@eumenes.ai. You also have the right to lodge a complaint with a supervisory authority, including: - your local EU data-protection authority (if you are in the EEA); - the UK Information Commissioner's Office (ICO) (if you are in the UK); or - the Personal Data Protection Commission (PDPC) of Singapore.
11.1. We use only strictly-necessary cookies required to operate the Service and keep you securely logged in — principally an authentication/session cookie and related device/enrolment identifiers. These cannot be switched off without breaking core functionality, and they do not require consent under applicable law because they are essential to provide a service you have requested.
11.2. We do not use advertising, analytics-for-advertising, or third-party tracking cookies, and we do not track you across other websites. Any telemetry we collect (Section 2.5) is first-party, used for operating and improving the Service, and is not used for advertising. On our public pages only, and only if you accept, we use Microsoft Clarity analytics (cookies _clck/_clsk) to understand how visitors use the site; it never runs inside your signed-in account. We use self-hosted, cookie-free analytics (Umami) on our public pages to count visits and sign-ups; no personal data or cookies, nothing shared with third parties.
11.3. You can control cookies through your browser settings, but disabling strictly-necessary cookies will prevent you from logging in and using the Service.
12.1. Where we rely on legitimate interests, you may object at any time on grounds relating to your particular situation, by contacting dpo@eumenes.ai. We will stop the relevant processing unless we have compelling legitimate grounds that override your interests, rights, and freedoms, or we need to process the data to establish, exercise, or defend legal claims.
12.2. Where we rely on consent (for example, voice processing, connecting accounts, or Support content), you may withdraw it at any time — for instance, by turning off the relevant feature, disconnecting the service, or deleting the content. Withdrawal does not affect the lawfulness of processing before withdrawal.
13.1. The Service is intended for users aged 16 or older and is not directed to children under 16. We do not knowingly collect personal data from children under 16. If you believe a child has provided us personal data, contact dpo@eumenes.ai and we will take steps to delete it.
14.1. We maintain measures designed to detect and respond to personal-data breaches. Where a breach is likely to result in a risk to your rights and freedoms (GDPR) or meets the notification threshold under the PDPA, we will notify the relevant supervisory authority and, where required, affected individuals, within the timeframes required by applicable law.
15.1. We may update this policy from time to time. If we make material changes, we will take reasonable steps to notify you (for example, by email or in-app notice) and will update the "Dated" line above. Your continued use of the Service after changes take effect constitutes acceptance of the updated policy, to the extent permitted by law.
For any privacy question, request, or complaint:
End of Privacy Policy (v1.0, 2026-08-28).