Voiso AI Code of Conduct

1. Purpose and scope

1.1 This AI Code of Conduct (the “Code“) sets the rules for using AI features of the Voiso platform, in particular AI Voice Agents. It forms part of, and is incorporated by reference into, the Voiso Terms & Conditions and any order form or master agreement between

Voiso Pte. Ltd., a Singapore company with its primary place of business at 9 Temasek Boulevard, #29-01, Suntec Tower 2, Singapore 038989, where the Customer is domiciled outside the USA or Canada, ]

and/or

Voiso Inc., a Delaware corporation with its primary place of business at 1000 N. West Street, Suite 1200, Wilmington, DE 19801, where the Customer is domiciled in the USA or Canada,

together with their subsidiaries and affiliates (“Voiso“, “we”) and the Customer (the “Agreement“). Breach of this Code is a material breach of the Agreement.

1.2 The Code applies to every Customer with AI Agent configured or run on a Customer account, whether inbound or outbound, and to every prompt, script, knowledge base, voice, routing rule, integration and API call used with it. It binds the Customer, its Authorized Users and, where the Customer is a reseller, BPO or agency, its own clients.

1.3 Customer control, Customer responsibility. Voiso provides the tooling and the Customer decides who the AI Agent talks to, what it says, what data it can see and what it may do. The Customer is therefore the party responsible, as against Voiso and any third party, for each AI Agent deployment and every interaction it conducts, regardless of whether a prompt or content was written by the Customer, its staff, its clients or with the help of AI.

1.4 Where this Code conflicts with the Agreement on any matter concerning AI Features, this Code prevails. Where a stricter rule applies under law or under the terms of a Third-Party AI Provider, the stricter rule applies.

2. Definitions

Capitalized terms not defined here have the meaning given in the Agreement.

Term	Meaning
AI Agent	Any AI-driven conversational agent configured on the Voiso platform, including AI Voice Agents for reception, support and outreach, and any chatbot using generative AI.
AI Features	AI Agents and all other AI functionality of the platform, including text-to-speech, transcription, call summaries, call scoring, speech analytics and AI answering machine detection.
Configuration	Everything the Customer supplies to shape an AI Agent: prompts, scripts, personas, greetings, workflows, routing and escalation rules, tools and integrations.
Customer Content	Configuration, Knowledge Base material, contact lists, and any other data the Customer submits, plus the recordings and transcripts of its calls.
Output	Speech, text, transcripts, summaries, scores, classifications or actions generated by AI Features.
Contact	Any natural person or business an AI Agent speaks to, whether they call in or are called.
Third-Party AI Provider	Any supplier of language, speech or voice models integrated into the platform.
High-Risk Use	Any use listed in section 5.3 or that falls in a high-risk category under applicable AI law, including Annex III of the EU AI Act.

3. Our commitments

Voiso commits to the following principles in how it builds and operates AI Features. These are statements of approach, not warranties, and do not expand Voiso’s obligations or liability under the Agreement.

1. We document what AI Features do, their known limitations and which Third-Party AI Providers they rely on, and we make available a default AI disclosure greeting that Customers can switch on.
2. Human handoff by design. AI Agents support warm transfer to a human agent, queue or number, so callers are never trapped in an automated loop.
3. Record-keeping. AI calls are recorded, transcribed, summarized and logged in the same CDR and reporting as human calls, so Customers can audit what was said where the local law allows such actions.
4. Privacy and security. AI Features run on the same infrastructure and security controls as the rest of the platform, described in the Voiso Trust & Compliance Center and Privacy Policy.
5. Vendor diligence. We review Third-Party AI Providers for security, privacy and legal risk before integrating them, and we do not permit them to train their models on Customer Content.
6. Safeguards and review. We may apply automated and manual checks to Configurations and usage to detect misuse (see section 12) and we update AI Features as law and technology change.

4. Customer responsibilities

4.1 Deployer status. For each AI Agent, the Customer acts as the deployer (and, where it places the agent on the market under its own name, the provider) for the purposes of the EU AI Act and similar laws. Voiso acts as a supplier of tooling and, for personal data, as processor on the Customer’s instructions.

4.2 The Customer must, for every AI Agent:

1. comply with all laws applicable to the agent’s calls, Contacts and content, including telecommunications, telemarketing, consumer protection, data protection, call recording, AI and sector-specific rules, in every jurisdiction where Contacts are located;
2. write, review, test and approve all Configuration before an agent takes a live call, and re-test after each material change;
3. ensure the Knowledge Base is accurate, current, lawful to use and free of third-party rights infringements;
4. hold every consent, licence, lawful basis and registration the agent’s activity requires, keep records of them, and produce them to Voiso in case of request;
5. monitor live performance, review a reasonable sample of recordings and transcripts, and correct errors without undue delay;
6. train its staff on this Code and keep Authorized Users’ credentials secure;
7. answer Voiso’s questions about the use case accurately and promptly, at sign-up and at any later time.

4.3 Representations. By activating an AI Agent, the Customer represents that it has the rights and consents needed to submit all Customer Content, to have Contacts’ calls processed by AI Features and Third-Party AI Providers, and to use the Output as intended.

4.4 Resellers, BPOs and agencies. A Customer that makes AI Features available to its own clients must bind them contractually to terms at least as protective of Voiso as this Code, vet their use cases, act promptly on violations, and, on request, identify which client is responsible for a given agent, number, campaign or content. The Customer is liable for its clients’ acts and omissions as for its own.

5. Prohibited and restricted uses

5.1 Prohibited uses. The Customer must not use, or allow an AI Agent to be used, to:

1. engage in or facilitate any unlawful activity, or violate sanctions or export control laws;
2. defraud or deceive Contacts, including phishing or vishing, fake prizes, invoices, tech support, investment, romance, “grandchild” or advance-fee scams;
3. impersonate any real person, company, public authority, bank, emergency service or other institution, or falsely claim to act on its behalf;
4. ask Contacts for passwords, PINs, one-time codes, full payment card data or government ID numbers, unless Voiso has agreed in writing and a compliant capture method is used;
5. harass, threaten, intimidate or abuse anyone, or run spam, “call bombing” or abusive robocall campaigns;
6. use manipulative techniques that exploit age, illness, disability, distress or financial hardship, or any practice prohibited under Article 5 of the EU AI Act;
7. infer emotions of employees from their voice, or use voice data for biometric identification or to infer sensitive characteristics, without the explicit consent the law requires;
8. target or market to minors, or create or transmit any content that sexualizes or exploits children;
9. produce hateful, violent, extremist, sexually explicit or self-harm content, or dangerous misinformation;
10. spread false information about elections or voting, discourage voting, or imitate a political candidate or official;
11. place automated calls to emergency or crisis numbers, generate artificial traffic, or run wangiri or other telecom fraud;
12. jailbreak, prompt-inject or otherwise circumvent AI safeguards, usage limits or monitoring;
13. reverse engineer the platform or models, extract prompts or Output to train competing models, or use synthetic voices to train speech or voice-cloning models.

5.2 High-Risk Uses. AI Agents must not give individual medical, legal, tax or investment advice, diagnose or triage medical urgency, act as the sole channel for emergencies, or take decisions based solely on automated processing that significantly affect a person (for example on credit, insurance, employment, housing or access to essential services). Where a use falls into a high-risk category under the EU AI Act or similar law, the Customer must meet all deployer obligations, including human oversight and informing affected persons, and must notify Voiso before going live.

6. Disclosure, consent and call recording

6.1 AI disclosure. Every AI Agent must tell Contacts they are speaking with an AI system, clearly and at the start of the conversation, unless that is obvious from the circumstances. An AI Agent must never claim to be human, deny being an AI when asked, or use a human name in a way intended to mislead. This reflects the deployer transparency duties under Article 50 of the EU AI Act, which have applied since 2 August 2026, and equivalent laws elsewhere. The Customer must not disable or shorten Voiso’s default disclosure greeting unless it has an equivalent disclosure in place.

6.2 Identification. On outbound calls, the AI Agent must state at the outset on whose behalf it is calling and, for marketing calls, the commercial purpose, and must give a callback number or other contact detail where the law requires.

6.3 Recording, transcription and analysis. AI Agent calls are recorded, transcribed and summarized by default. The Customer is solely responsible for informing every participant and obtaining any consent that one-party or all-party recording laws, data protection law or sector rules require, and for configuring recording off where consent cannot lawfully be obtained.

6.4 Human on request. Where a Contact asks to speak to a human, the AI Agent must offer a transfer, a callback or another route to a person, unless the Customer has clearly told the Contact at the start that no human is available for that line.

6.5 Synthetic content. Where law requires AI-generated audio or text to be labelled or marked as synthetic, the Customer must ensure that labelling is applied to any Output it publishes or forwards.

7. Telemarketing and outbound calling

7.1 Consent first. The Customer may only use an AI Agent to call or message a Contact where it holds the consent or other lawful basis the law requires. In the United States, the FCC confirmed in its February 2024 Declaratory Ruling that AI-generated voices are “artificial” voices under the TCPA, so AI Agent calls require the same prior express (and, for marketing, written) consent as prerecorded calls. Equivalent rules apply in the EU, UK and elsewhere, including ePrivacy, PECR and national unfair-competition laws.

7.2 Lists. The Customer is solely responsible for the origin and lawfulness of every contact list it uploads, and must not dial randomly generated, sequential or purchased numbers whose consent it cannot evidence. Voiso’s Number Validator, AI answering machine detection and similar tools do not establish consent.

7.3 Do-not-call and opt-outs. The Customer must scrub lists against applicable do-not-call and Robinson registers, configure AI Agents to accept opt-out requests made during a call, and honour them promptly and permanently across all its campaigns.

7.4 Times and frequency. Calls must be placed only at hours permitted in the Contact’s time zone and at a reasonable frequency.

7.5 Caller ID. The Customer must display a number it is assigned or authorized to use and at which it can be reached. Spoofing, unlawful suppression, or rotating numbers to evade blocking, spam labels or complaints is prohibited. Local Caller ID numbers must be used in line with the rules of each country and number range.

7.6 Lead validation calls. Outreach agents that validate leads or confirm interest are marketing calls for these purposes unless the Customer has established otherwise, and must meet sections 6 and 7 in full.

7.7 Carrier charges. Any fines, chargebacks, penalties or blocking costs imposed on Voiso by carriers, number-range holders or regulators because of the Customer’s calls are recoverable from the Customer.

8. Voices, voice cloning and impersonation

8.1 The Customer may use only voices from the Voiso voice library or voices it has the right to use.

8.2 The Customer must not clone or imitate any real person’s voice unless that person has given explicit, informed, written (or equally durable and verifiable) consent to the specific use. The Customer must keep that evidence, provide it on request, stop using and delete the voice without undue delay if consent is withdrawn, and obtain a parent’s or guardian’s consent for a minor.

8.3 The Customer must never use synthetic voices to imitate celebrities, public figures, political candidates or office holders, or deceased persons; to bypass voice-based authentication or voice biometrics; or to attribute statements to anyone they did not make.

8.4 Agent personas are permitted, provided the persona is disclosed as an AI under section 6.1 and is not presented as a real employee.

9. Data, privacy and confidentiality

9.1 Roles. For personal data processed by AI Features, the Customer is the controller and Voiso a processor under the Voiso Data Processing Agreements. The Customer is responsible for having a lawful basis, giving Contacts the privacy information the law requires (including that calls are handled, recorded and analysed by AI and by Third-Party AI Providers), and handling data subject requests.

9.2 Minimization. The Customer must not load into prompts or Knowledge Bases more personal data than an agent needs, and must not instruct AI Agents to collect special-category data (health, biometric, religious, political, sexual-life data and similar), children’s data, or financial and authentication credentials unless it has a lawful basis, Voiso’s prior written approval and appropriate safeguards. Protected health information may only be processed under a signed business associate agreement.

9.3 Third-Party AI Providers. AI Features use third-party language, speech and voice models, listed in Voiso’s web-site. The Customer authorizes Voiso to share Customer Content with them as needed to deliver the AI Features, and must comply with any provider usage policies Voiso makes available. If a provider restricts a use, Voiso may restrict the related feature.

9.4 Model training. Voiso will not use Customer Content to train general-purpose models made available to other customers without the Customer’s consent. Voiso may use de-identified, aggregated usage data and Output metadata to operate, secure, and improve the AI Features.

9.5 Confidential information. The Customer must not place in Configuration or Knowledge Bases any information it is not entitled to disclose, and should assume an AI Agent may repeat Knowledge Base content to any Contact who asks. The Customer must not include secrets, internal-only pricing, credentials or third-party confidential information it does not want disclosed.

10. Human oversight, escalation and actions

10.1 The Customer must keep a human able to oversee each AI Agent, review its outcomes and stop it. It must configure warm transfer or another escalation route for: callers who ask for a person; complaints; vulnerable callers; anything outside the agent’s defined scope; and any statement of an emergency, risk to life or self-harm, which must be directed to emergency services or a human immediately.

10.2 Agent actions. Where an AI Agent is connected to tools, CRMs or APIs that let it change records, book, cancel, refund, take payments or commit the Customer, the Customer chooses those permissions and is bound by every action the agent takes within them. The Customer should require human confirmation for irreversible or high-value actions.

10.3 Commitments to Contacts. Statements, offers, prices, promises or agreements made by an AI Agent are made by and on behalf of the Customer, not Voiso. The Customer must configure agents so they do not make commitments it is not prepared to honour.

10.4 Pre-launch testing. Before going live and after each material change, the Customer must test each AI Agent using the Voiso live-test tool on realistic scenarios, including attempts to make it go off-script, disclose confidential data or act outside its scope.

11. Limitations of AI Output

11.1 Probabilistic output. The Customer acknowledges that AI Features are probabilistic and may produce Output that is inaccurate, incomplete, inconsistent, biased, offensive or not grounded in the Knowledge Base (“hallucinations”); may misunderstand speech, accents or intent; may mis-transcribe or mis-summarize calls; and may misclassify answering machines, sentiment or call scores. Response times and performance vary with call complexity, language, network and geography.

11.2 No reliance. Output is not professional advice and is not a statement by Voiso. The Customer is responsible for evaluating Output before relying on it, and must not use Output, scores or summaries as the sole basis for decisions with legal or similarly significant effect on any person, including staff performance or disciplinary decisions.

11.3 As-is. To the maximum extent permitted by law, and without limiting any express service levels in the Agreement, AI Features and Output are provided “as is” and “as available”, and Voiso disclaims all warranties that Output will be accurate, lawful for the Customer’s purpose, error-free or uninterrupted, or that any AI Feature will achieve a particular result, conversion rate or cost saving.

11.4 Third-party models. Voiso does not control the behaviour, availability or changes of Third-Party AI Provider models, and may substitute, update or withdraw models, voices or languages. Voiso will try to give reasonable notice of material changes.

12. Monitoring, enforcement and suspension

12.1 Review. Voiso may, to the extent necessary and in line with data protection law, use automated and manual means to review Configurations, greetings, Knowledge Bases, call metadata, recordings and transcripts, in order to enforce this Code, protect Contacts, networks and the platform, and meet legal obligations. Voiso may automatically check Configurations when saved and block an agent that shows a serious risk, such as impersonation, a missing AI disclosure, spoofing or fraud. Voiso has no obligation to monitor and is not responsible for Customer Content.

12.2 Information. The Customer must cooperate with any investigation, and provide within reasonable time the information Voiso reasonably requests, including evidence of consent, list sources and use-case descriptions.

12.3 Measures. If Voiso reasonably believes this Code has been or is likely to be breached, or that an AI Feature is being used in a way that creates legal, regulatory, reputational or network risk, Voiso may, proportionately and without liability: issue a warning; require remediation; disable an agent, campaign, number or Knowledge Base; reduce limits or restrict features; suspend AI Features or the account; or terminate the Agreement for cause. Fees for the affected period remain payable.

12.4 Urgent cases. Where there is ongoing fraud, risk to people, a regulator or carrier demand, or serious harm to numbers, networks or Voiso’s providers, Voiso may act immediately without notice and will inform the Customer as soon as reasonably practicable, unless legally prohibited.

12.5 Authorities. Voiso may report suspected unlawful activity to competent authorities and carriers, and cooperate with traceback, law enforcement and regulatory requests.

12.6 Appeal. A Customer may ask for a manual review of any measure by writing to [email protected]. Voiso will reverse a measure without undue delay if it was unjustified. No waiver arises from Voiso not acting on a breach.

13. Liability and indemnity

13.1 Indemnity. The Customer will defend, indemnify and hold harmless Voiso, its affiliates, officers, employees, and Third-Party AI and telecom providers from and against all claims, losses, damages, fines, penalties, carrier charges, regulatory costs, settlements and reasonable legal fees arising out of or relating to: (a) any breach of this Code; (b) Customer Content, including Configuration and Knowledge Bases; (c) any call, message, statement, commitment or action made by an AI Agent on the Customer’s account; (d) the Customer’s failure to obtain any consent, licence or lawful basis, or to give any required disclosure; and (e) the acts and omissions of the Customer’s Authorized Users and clients.

13.2 Exclusion. To the maximum extent permitted by law, Voiso is not liable for any loss arising from Output, from reliance on Output, or from an AI Agent’s conduct, except to the extent caused by Voiso’s breach of the Agreement, wilful misconduct or gross negligence.

13.3 Mandatory law. Nothing in this Code limits liability that cannot be limited under applicable law, including for death or personal injury caused by negligence, or for fraud.

14. Reporting, changes and contact

14.1 Report misuse. Anyone, including a person who received a call from a Voiso-hosted AI Agent, can report suspected misuse to [email protected], giving where possible the calling number, the number called, date, time and time zone, and what happened. Voiso will acknowledge reports that include contact details and act on them in line with section 12.

14.2 Security issues. Vulnerabilities, prompt-injection paths or safety failures in AI Features should be reported confidentially to [email protected] and not tested on the live platform without written permission.

14.3 Changes. Voiso may update this Code to reflect changes in law, regulator or carrier requirements, Third-Party AI Provider terms or AI Features. Material changes will be notified 30 days in advance by email or in the admin, except where faster change is required by law or to address a risk of harm. Continued use of AI Features after the effective date means acceptance. The current version will be published at voiso.com/ai-code-of-conduct.

14.4 Severability and language. If any provision is unenforceable, the rest remains in effect. The English version prevails over any translation.

14.5 Contact. Questions about this Code: [email protected], Voiso Pte. Ltd., 9 Temasek Boulevard, #29-01, Suntec Tower 2, Singapore 038989 (Customers outside the USA and Canada); Voiso Inc., 1000 N. West Street, Suite 1200, Wilmington, DE 19801 (Customers in the USA and Canada).