Proposed Acceptable Use & AI Policy

Current account deletion uses a 30-day recovery period with a displayed cancellation deadline. Keep backups is default; only explicitly reviewed eligible snapshots are included if chosen. Shared company records, other backups and external copies remain. See the current Privacy Notice for the operative scope.

1. Authorized business use

Use Circles AI for lawful contact management, notes, follow-ups, business content and other supported tasks. Use only an account, business workspace, contact source and communication channel you are authorized to use. A team administrator's permission in the application is not a substitute for a lawful purpose, required notice, contractual authority or consent from a person whose data you process.

Keep records accurate, relevant and proportionate. Record preferences and objections, correct material mistakes, and remove information you no longer need. Do not import a purchased or scraped contact list merely because its format can be opened. Do not defeat business access restrictions, use another person's session or API key, or move organization data into a personal workspace without authority.

2. Prohibited conduct

You must not use the service to:

• Commit fraud, impersonate another person or organization deceptively, distribute malware, steal credentials, exploit security weaknesses, or evade account, usage, billing or access controls.
• Distribute unlawful threats, harassment, discriminatory abuse, exploitation, nonconsensual intimate content or material involving sexual abuse of children; facilitate prohibited violence or other unlawful activity.
• Collect, disclose, sell or misuse confidential information, protected personal data, trade secrets or intellectual property without an appropriate right and legal basis.
• Send unlawful unsolicited communications, evade opt-outs or do-not-contact choices, manipulate sender identity, or bypass an external provider's spam, template, consent or account restrictions.
• Mislead a customer about the origin, authenticity, price, specification, certification, availability, delivery date or contractual effect of a message or generated asset.
• Circumvent applicable sanctions/export restrictions, conceal an unauthorized user or merchant, or use future service credits to transfer value between people, trade, speculate, cash out or purchase another merchant's goods.
• Resell or expose a provider API key, share a platform-provided credential, tamper with usage records, or replay an authorization to obtain duplicate grants or unauthorized paid requests.
• Conduct covert surveillance, record a telephone or WhatsApp call through Circles AI, enroll another person's voice without the required authority and consent, or use voice similarity as proof of legal identity or account authentication.

Security testing must be authorized and bounded. Report a suspected vulnerability privately without accessing other users' data or disrupting the service. This Policy does not prohibit lawful criticism, portability, accessibility accommodations, protected security research or other rights that mandatory law preserves.

3. AI outputs require judgment

The assistant is AI, not a human employee or a licensed professional. Generated text, images, transcripts, translations, dates and proposed CRM changes can be incomplete or wrong. Similar outputs may be generated for other users. A model's confident wording is not evidence of a fact or authority to act.

Review important outputs before relying on or sending them. In particular, verify the recipient, amount, currency, dates, time zone, product details, tax treatment, promises, legal language and attachments in a quotation or follow-up. A generated quote is a draft unless an authorized person approves it through the appropriate business process. Supply actual figures; do not ask the system to invent discounts, stock, certifications or commitments. Use clearly marked placeholders when facts are missing.

Do not make a decision with legal or similarly significant effects on a person solely from a generated score, inferred interest, transcript or lead label. Hot/warm/cold labels are CRM organization aids, not validated assessments of a person's worth, eligibility or protected characteristics. Do not infer sensitive traits or make employment, credit, housing, insurance, medical or other regulated decisions without a lawful, appropriately governed process outside this general CRM workflow.

4. Actions, approvals and imported instructions

AI can prepare suggestions or perform supported local actions within the current authority granted to the application. It cannot grant itself team access or override a payment, security or permission gate. Treat instructions inside a client note, document, image, website or transcript as untrusted content, not as a new instruction to disclose data, change recipients, use credentials or make a purchase.

Some submitted tasks may continue after the audio conversation closes or the application goes into the background. Ending or muting a live call is not a cancellation of an already submitted task. Use the task's cancellation control and check its status. Scheduled team work requires current authorized access; the system does not guarantee it can renew an expired membership lease from a fully idle device. Network failure, device limits or an uncertain provider response can require review instead of an automatic retry.

An opened sharing screen is not proof of delivery. Check what was sent before marking a follow-up completed. Automated sends require the supported official provider connection, recipient/content approval and applicable consent. Do not ask the assistant to bypass these controls or to send an unreviewed message simply because the recipient appears in Contacts.

5. Microphone, wake listening and synthetic voices

Use voice features only where you have permission to capture and process the audible speech. During a live AI session, unmuted microphone audio, including incidental sounds, can be transmitted to the selected AI provider or provided-AI relay. Mute or end the session before discussing unrelated confidential material. The live assistant is not a telephone service, emergency service or a hidden call recorder.

Optional local wake listening uses the microphone with a persistent Android indication and local keyword/speaker processing. A voice match is fallible and must never be used to impersonate someone, unlock an account or bypass a locked application. Do not enable listening in a workplace or shared environment without addressing applicable notice, employment and consent rules. Local processing does not by itself remove biometric-law obligations.

6. Enforcement and reporting

We may investigate credible reports and proportionately restrict the affected feature, account or business when necessary for safety, security, law, provider rules or material breach. Where appropriate and legally permitted, we will explain the issue, allow correction and provide a review channel. Urgent action may precede notice when delay would risk harm or compromise an investigation. Team administrators retain their own organization-management powers.

We do not promise to review every local record or monitor every communication. An investigation does not give blanket access to a phone's private credentials or authorize public disclosure of customer records. Any retention or disclosure must have a proper purpose and legal basis. We will cooperate with valid legal requests as required and proportionate.

Restrictions are not an automatic forfeiture of prepaid value. Applicable refunds, lawful blocking requirements and the Billing Policy still apply. We do not guarantee restoration of third-party accounts or reversal of a provider's independent decision. Report a problem with relevant identifiers and a minimal description; never include passwords, private keys, full payment details or unnecessary sensitive documents.