Legal
Acceptable Use Policy
Last updated: November 9, 2026
This Acceptable Use Policy (“AUP”) governs Customer’s and its Authorised Users’ access to and use of Eloquent AI’s Services, including AI operators, software, APIs and integrations. It forms part of the Eloquent AI Terms of Service (“Agreement”). Capitalised terms not defined here have the meanings given in the Agreement. References to “Platform Services” mean the Services.
The order of precedence in the Agreement applies to this AUP. The DPA and mandatory international-transfer instruments retain their applicable precedence.
Changes to this AUP are governed by Section 8.4 of the Agreement.
Data Protection. When you use the Platform Services to process personal data for your end-users, you act as Controller and Eloquent AI acts as Processor in accordance with our Data Processing Addendum (DPA). The DPA forms part of your agreement with us and prevails to the extent necessary to comply with applicable data protection laws.
For safety, compliance and platform integrity, Eloquent AI may monitor usage for abuse detection, troubleshooting and policy compliance, subject to the Agreement, confidentiality obligations and the DPA. Personal Data processed on Customer’s behalf remains subject to documented instructions and the DPA, including its retention and deletion requirements. Independent-controller processing is governed by the Privacy Policy. Monitoring does not authorise unrelated use, unrestricted model training or indefinite retention. Suspension, restriction and termination are governed by the Enforcement section below.
Eloquent AI may also implement reasonable technical safeguards, filtering mechanisms, rate limits, usage thresholds, access restrictions, or protective controls designed to maintain the security, integrity, availability, compliance, or stability of the Platform Services and related systems.
Prohibited Activities
To maintain the integrity, security, and proper use of Eloquent AI’s Platform Services, you agree not to:
1. Interfere with the Platform Services:
- Modify, reverse engineer, decompile, disassemble, decrypt, or otherwise attempt to access the human-readable form of the Platform Services.
- Create adaptations, translations, or derivative works based on the Platform Services.
- Undertake platform benchmarking, load testing or performance testing, or publish comparative results, without prior written approval. Ordinary non-intrusive functional validation, agreed workflow simulation and internal quality review of your authorised instance are permitted under the Limited Exception for Testing below.
- Attempt to bypass or defeat model safety features, jailbreak protections, or other safeguards.
- Perform penetration testing, vulnerability scanning, adversarial testing, prompt injection testing, red-team activities, or security assessments against the Platform Services without Eloquent AI’s prior written approval.
2. Circumvent Usage Restrictions:
- Bypass or attempt to bypass user limits or usage restrictions built into the Platform Services.
- Remove any proprietary notices, branding, or ownership labels from Eloquent AI’s technology.
- Circumvent authentication, rate limits, throttling, or usage caps.
3. Unauthorised Sharing or Distribution:
- Disseminate, market, sublicense, rent, lease or lend the Services, or operate a service bureau, without Eloquent AI’s written consent. Authorised customer-facing interactions through the websites, applications, chat, voice, email and social channels included in Customer’s selected subscription plan, accepted online order or applicable Order Form, and authorised integrations under the Agreement and API Terms, do not constitute prohibited sharing or resale. This exception does not permit third parties to resell the Services or access administrative accounts or APIs without authorisation.
4. Tampering and Unauthorised Access:
- Interfere with, disrupt, destroy, or unlawfully access server networks, data, or other assets belonging to Eloquent AI or its partners.
- Use the Platform Services to gain unauthorised access to third-party systems, violate third-party terms of service, bypass access controls, or engage in abusive automated interactions with external platforms or services.
- Access, process, monitor, or automate any systems, accounts, communications, or data without all necessary rights, authorisations, notices, and consents required under applicable law, contract, or organisational policy.
- Use the Platform Services in restricted or sanctioned countries or territories or access features where such use is prohibited.
Upload malware, malicious prompts, or other content intended to degrade security or performance.
5. Unethical Use:
- Use Eloquent AI’s proprietary technology or service-generated material to build competing products, copy protected platform features or replicate the Services.
Use the Services, outputs, workflows, operational behaviour, or data generated through the Services to:
- train, fine-tune, evaluate, benchmark or improve any competing artificial intelligence, machine-learning, automation or operator product;
- create datasets, synthetic datasets, or derivative training materials for use in competing products or services; or
- replicate or attempt to replicate Eloquent AI’s proprietary functionality, behaviour, workflow implementation or operational logic in a competing product.
- Input malicious prompts or attempt to manipulate the Services to generate harmful, illegal or offensive content, reveal sensitive or confidential information without authority, or bypass approved instructions and safeguards. Authorised testing is subject to the Limited Exception for Testing below.
Customer retains ownership of its input data, internal business rules, policies, procedures and independently developed workflows. The competitive-use restrictions do not independently restrict lawful reuse of those materials, provided such reuse does not use Eloquent AI’s proprietary technology, outputs or observed platform behaviour to replicate or improve a competing product. Authorised internal quality evaluation and configuration of the purchased Services are permitted.
6. Improper Use of Generative AI Features:
- Generate medical, legal, financial, immigration or other regulated professional advice, or portray outputs as such. The Services do not themselves provide regulated professional advice. This restriction does not prohibit factual customer support, information retrieval, approved record updates, repayment-date changes, screening or administrative assistance within the agreed scope and safeguards. It does not authorise autonomous credit, underwriting or other consequential decisions.
- Produce content that could lead to biased or unfair treatment of individuals or protected classes.
- Make solely automated decisions that produce legal or similarly significant effects on an individual without appropriate human review, transparency, and a lawful basis. Customer is responsible for ensuring that any legally required notices, disclosures, consents, or transparency obligations relating to the use of artificial intelligence or automated systems are properly implemented and maintained.
- Use the Services as the sole basis for decisions requiring legally mandated human oversight, professional judgment, regulatory approval or licensed professional review. Customer remains responsible for its business, compliance, legal, underwriting, fraud, payment, servicing, customer-support and regulatory decisions, including legally required oversight and review. Written approval by Eloquent AI or inclusion of a workflow in an Order Form never waives applicable law or a required approval or professional review.
- Use the Platform Services in situations where failure, inaccuracy, delay, or unavailability could reasonably be expected to result in death, bodily injury, emergency response failures, or other critical safety harms unless expressly approved by Eloquent AI in writing.
- Mislead others by claiming generative AI outputs were created by humans; create synthetic/edited media (deepfakes or voice clones) without clear labelling and all required consents.
- Use the Services to impersonate individuals or organisations, conduct phishing or social engineering, misrepresent an automated system as human, or engage in fraudulent, deceptive or misleading conduct.
High-Risk and Harmful AI Uses
Do not use the Services for:
- unlawful surveillance, social scoring, or discriminatory profiling;
- exploitation of vulnerable individuals or groups;
- generation or dissemination of knowingly false or misleading information intended to cause harm;
- unlawful biometric identification or processing;
- activities involving child exploitation, including AI-generated child sexual abuse material or grooming, human trafficking, terrorism or violence;
- development or deployment of malicious code, malware, or cyberattack capabilities; or
- promotion or facilitation of suicide, self-harm or disordered eating; creation or distribution of non-consensual intimate content or sexual violence; or any use prohibited under applicable artificial intelligence, consumer-protection, financial-services or data-protection laws.
7. General Misuse:
- Engage in activities that infringe on Eloquent AI’s intellectual property rights or those of third parties.
- Violate any applicable laws, including export control and sanctions regulations.
- Use the Platform Services for the benefit of any individual, entity, organisation, or jurisdiction subject to applicable sanctions, export restrictions, anti-money laundering laws, or trade compliance prohibitions imposed by the United States, United Kingdom, European Union, or other applicable governmental authority.
- Use the Services in a manner that causes Eloquent AI or its infrastructure, cloud, AI or model providers to violate applicable laws, regulations or provider usage restrictions applicable to the purchased features and identified or made available to Customer. Eloquent AI will identify relevant restrictions through documentation, the Order Form or written notice; material changes follow the applicable contractual notice process. Urgent restrictions compelled by an upstream provider or binding law may be addressed only through protective measures permitted by the Agreement, with notice as soon as reasonably practicable.
- Conduct fraudulent, harmful, deceptive, defamatory, harassing or abusive activities, including abusive debt collection, deceptive repayment demands, fabricated legal or regulatory claims, or manipulative sales or collection practices exploiting financially vulnerable individuals.
- Compromise network security, crack passwords, or circumvent encryption.
- Send unsolicited or unlawful communications, including spam or non-compliant telemarketing messages.
- Automate processes, such as spam or unauthorised data scraping, that interfere with the Platform Services’ performance.
- Deploy or configure the Platform Services in a manner that creates uncontrolled, recursive, self-replicating, or autonomous execution loops that may materially impact platform stability, security, performance, or third-party systems.
8. Restricted Data and Credentials
Routine customer operations. Customer may provide or permit access to ordinary customer data, including names, contact details, account information and service conversations, as necessary for the agreed workflows and subject to the Agreement and DPA. This permission does not extend to the restricted categories below.
Planned sensitive-data workflows. Planned collection or use of the following categories requires express authorisation in an Order Form or amendment to the Agreement signed by both parties or a completed customer-specific DPA annex identifying this AUP provision that is signed by both parties or attached to and expressly incorporated into an Order Form signed by both parties, and is permitted only with the agreed safeguards. Limited handling of unexpected disclosures is addressed below:
- Special-category/sensitive data (e.g., health, biometric/faceprints/voiceprints, precise geolocation, union membership, racial/ethnic origin), children’s data, or criminal-records data;
- Payment card data beyond truncated PAN or tokenised values;
- Government-issued identifiers, including full passport numbers, SSNs or NINOs. Planned collection or use requires written authorisation and must be strictly necessary and lawfully justified; necessity alone does not grant permission;
- Credentials and secrets, including seed phrases, private keys, API client secrets, production passwords and SSH keys, except credentials supplied through the approved secure authentication and integration mechanisms described below;
- Data subject to ITAR, classified controls, or other export-restricted regimes.
An authorisation must specify the permitted categories, processing purpose, workflow, approved environment, access controls, retention, duration and any additional safeguards. It covers the agreed workflow and data categories; separate approval is not required for each end-user interaction within that scope. Material changes to the authorised scope require further written approval. Record the customer-specific authorisation and processing details in the applicable DPA annex; use Annex 4 for passport onboarding and related workflow recordings. The Order Form need only incorporate the completed annex. A general reference to sensitive data or account credentials in a policy or DPA does not itself authorise restricted inputs. No authorisation permits unlawful processing or prohibited export-controlled activity.
Unexpected disclosures. An end user’s unsolicited disclosure of restricted data during an otherwise authorised interaction does not, by itself, constitute a breach of this AUP. Customer must use available controls and agreed procedures to avoid further solicitation, restrict access, minimise copying and retention, and arrange lawful redaction or deletion or secure human escalation as appropriate, subject to the DPA and applicable law. Handling is permitted only to the extent necessary and lawful to contain and appropriately route the disclosure. If such disclosures recur or become reasonably foreseeable, Customer must work with Eloquent AI to review safeguards and obtain authorisation for any intended ongoing processing. This exception does not permit unrelated use, unlawful possession, unapproved systematic collection or circumvention of provider restrictions.
API keys, OAuth tokens, client credentials and other necessary integration credentials may be provided through authentication fields, secrets-management facilities or secure integration channels expressly approved by Eloquent AI, solely for an authorised integration and subject to the API Terms. Use least-privilege permissions, restrict access, rotate or revoke compromised credentials promptly and report suspected compromise. Do not paste credentials into prompts, conversations, knowledge bases, ordinary file uploads or support logs. This permission does not extend to unnecessary financial seed phrases or private keys, or to unapproved collection of an end user’s credentials.
For processing, security, subprocessors and international transfers, see the DPA and Subprocessor List.
Required Deployment Safeguards
- Approved scope and authority. Before deployment, Customer must approve and document each workflow’s purpose, permitted actions, connected systems, data access, authorisation limits, human approval points and escalation criteria. Record material use-case restrictions and approval thresholds in the applicable Order Form, accepted online order or documented workflow approval. Material changes must be assessed and revalidated before production use.
- Autonomous-action controls. Apply least-privilege access, risk-appropriate action and transaction limits, monitoring, anomaly detection, failure alerts and effective means to pause or disable automation. Sensitive or irreversible actions require risk-appropriate approval gates. Customer must maintain controls within its responsibility and cooperate with Eloquent AI on controls allocated to Eloquent AI under the Agreement.
- Consequential decisions. Workflows materially affecting credit eligibility, underwriting, insurance coverage or claims, access to essential services, or consequential fraud decisions require a documented assessment, approval by an authorised Customer owner and review by appropriately qualified personnel before a consequential decision is finalised or communicated. Provide a human escalation and correction route. Routine factual servicing and approved administrative tasks may be automated within the agreed safeguards; this does not authorise autonomous consequential decisions. Any stricter applicable provider restriction also applies.
- AI disclosure and consent. Customer is responsible for preparing, configuring and maintaining AI-interaction disclosures, call-recording notices and any legally required consent process. Required AI disclosures must clearly identify the interaction as involving an AI operator, including a synthetic voice, no later than the first interaction. Recording notices and consents must be provided or obtained before recording where required by law. Customer must obtain all legally required communication and synthetic-voice consents, repeat disclosures where necessary to avoid misleading users, respect applicable opt-outs and preserve required labelling or provenance information. Customer must not disable or bypass legally required notices or consent controls. Each party remains responsible for obligations that applicable law places directly on it.
- User reports and remediation. Provide an accessible route for end users to report harmful outputs, incorrect actions or misuse. Respond reasonably, investigate material issues, promptly contain harmful automation within Customer’s control, cooperate with remediation and escalate unresolved risks to Eloquent AI.
- Authorised users. Customer is responsible for its authorised personnel and systems and must take reasonable measures to communicate and enforce applicable restrictions in customer-facing deployments. Do not permit unauthorised third-party access or use alternative accounts to evade a suspension.
Controlled Handling of Harmful Content
Receiving or processing genuinely incoming harmful content solely for an authorised customer-support, moderation, investigation or reporting workflow is not, by itself, prohibited use. Limit access, copying, retention and onward disclosure to what is necessary and lawful; use appropriate filters and human escalation; and do not promote, exploit, republish or amplify the harmful material. Follow any applicable mandatory reporting or handling requirements and provider restrictions. This provision does not permit intentional generation of prohibited content, unlawful possession, or use of otherwise restricted personal data beyond the authorisation or limited unexpected-disclosure handling permitted by Section 8. Intentional safety testing remains subject to the written approvals below.
Limited Exception for Testing
You may perform ordinary non-intrusive configuration, functional validation, agreed workflow simulation and internal quality review of your authorised instance for legitimate internal business purposes. This is not a general exemption from the AUP. Such activities are permitted only within the agreed scope and provided that:
- Such activities do not materially impact the security, integrity, availability, performance, or lawful operation of the Platform Services or third-party systems;
You promptly report to Eloquent AI any:
- security vulnerabilities;
- harmful or illegal content;
- unauthorised exposure of third-party confidential information.
Penetration testing, vulnerability scanning, adversarial or prompt-injection testing, red-teaming, security assessments, platform benchmarking, load testing and publication of comparative results require Eloquent AI’s prior written approval defining scope, timing, environment and safeguards, and any additional approval required by the applicable provider. Do not test third-party systems without their owner’s permission. DPA audit rights remain available on their agreed terms, but do not independently authorise intrusive testing.
Use synthetic or appropriately de-identified test data. Do not put identifiable confidential Customer Data in non-production environments contrary to the Agreement or applicable data-handling policy; any exception requires the necessary written approval and safeguards. Testing does not override restricted-data, confidentiality, legal, retention, destruction or reporting requirements. Prohibited-content generation requires express written approval for a lawful safety-testing purpose and remains subject to applicable law and provider restrictions.
Reporting Abuse
Report abuse, harmful outputs and policy concerns to privacy@eloquentai.co. For urgent Priority-1 security incidents, including credential compromise or active data exposure, use emergency.support@eloquentai.co in accordance with the Terms of Service. Provide timestamps, request IDs and only the minimum necessary redacted logs. Do not include passwords, API secrets, unnecessary personal data or unlawful content in an ordinary email; request an approved secure channel where sensitive evidence is necessary. These operational contacts do not replace contractual notice procedures in Section 8.3 of the Terms of Service.
Enforcement
Eloquent AI may investigate reasonably suspected violations and temporarily suspend, restrict or throttle affected Services where reasonably necessary to prevent or address material security risks, fraud, unlawful activity, material misuse or material breach. A suspected violation must have a reasonable factual basis; suspicion alone does not create a separate termination right. Suspension, restriction and related notice requirements are governed by Section 2.4.1 of the Terms of Service. Cure periods and termination are governed by Section 7.
Eloquent AI will use commercially reasonable efforts to limit the scope and duration of restrictions and restore access promptly once the underlying issue is resolved. Customer must reasonably cooperate with investigations, stop material misuse, lawfully preserve relevant evidence and implement proportionate remediation. Customer may request review at support@eloquentai.co and provide supporting information. Eloquent AI will review the request in good faith and explain its outcome to the extent legally permitted and consistent with security and confidentiality.
Termination, including applicable cure periods, is governed by Section 7 of the Terms of Service. This AUP creates no additional right to terminate for mere suspicion and does not override agreed fees, refunds, availability remedies, data-return or deletion obligations. Eloquent AI may take immediate protective action only as permitted by the Agreement or required by applicable law.
Eloquent AI may take reasonable technical, operational, or legal measures necessary to protect the security, integrity, availability, or lawful operation of the Platform Services, its customers, third-party systems, or the public.
Survival
Provisions that by their nature remain relevant after termination, including restrictions on misuse of retained Eloquent AI intellectual property and obligations concerning lawfully retained evidence, survive to that extent. Accrued rights and remedies are preserved. Confidentiality, intellectual property ownership, indemnification and liability limitations survive as provided in the governing Agreement. Survival does not extend access rights, create new liabilities or permit retention or use of Customer Data beyond the Agreement and DPA.
Contact
Eloquent AI, Inc.
456 Montgomery Street, Suite 1450, San Francisco, CA 94104, United States
privacy@eloquentai.co