AI Features Addendum

Supplement to the OwnSig EULA and Privacy Policy.

This Addendum governs OwnSig's optional AI features (the assistant "Penna", AI signature generation, AI image generation, and related capabilities). AI features are off by default and only operate after an organisation administrator explicitly enables them and accepts this Addendum. Using OwnSig without AI features does not require acceptance of this Addendum.

1. Opt-in

• AI features activate only when a tenant administrator enables them under Admin → AI assistant and records acceptance of this Addendum. Acceptance is timestamped and attributed.
• Administrators can disable AI features at any time; processing under this Addendum stops with them.
• Feature controls allow restricting AI features to specific roles inside the tenant.

2. What the AI features process

• Prompts you type to Penna, signature field content submitted for generation or improvement, and image-generation prompts.
• Workspace data returned by Penna's permission-checked tools (only what the requesting user could already see).
• AI features never receive message bodies of your email, recipient identities, or tracking events beyond aggregate counts — except employee sentiment analysis (Section 2a), which is a distinct, separately-consented capability.

2a. Employee sentiment analysis (separate opt-in)

Sentiment analysis is a distinct add-in capability, off by default, that an administrator must enable separately from the base AI features above (Admin → Employee guidance) and separately accept these terms for. When enabled:

• The text of the current drafta person is composing in Outlook/Gmail is sent to the organisation's configured AI provider (BYOK or managed) to produce a tone score. This is the one AI capability that processes draft content.
• Only the derived score, category, and (per the administrator's configured granularity — organisation-wide, by department, or per employee) an identifier at that granularity are stored. The draft text itself is never stored, logged, or used for training, on any plan.
• An administrator can configure alerts that notify designated recipients when a pattern of low scores recurs for an employee or department, per the granularity enabled.
• Works-council / co-determination mode (Admin → Policies) disables per-employee granularity and per-employee alerting regardless of this setting — only aggregate reporting remains available.

3. Model improvement (training)

• Free, Individual, and Pro plans: by accepting this Addendum, you agree that AI feature inputs and outputs (prompts, generated signatures, feedback signals) may be used by OwnSig to train and improve its AI features and service quality. We de-identify this data before use, and we never use it to build profiles of your recipients.
• Enterprise and Enterprise Sovereign plans: tenant content is excluded from training. AI inputs and outputs are processed transiently to provide the feature, retained only per the operational log policy, and never used to train or improve models. This exclusion is automatic — no configuration required — and survives in your MSA.
• Bring-your-own-key (BYOK) requests go to your chosen provider under your agreement with that provider; OwnSig does not train on BYOK traffic in transit.

4. Sub-processors

Managed AI routes requests to contracted model providers under data-processing terms that prohibit them from training on your content. The current provider list appears on the security page.

5. Your responsibilities

• Review AI output before deploying it — generated content can be inaccurate.
• Do not submit special-category personal data or third-party confidential information in prompts.

6. Changes

Material changes to this Addendum (including any change to the training boundary) require re-acceptance by a tenant administrator before AI features continue.