Service Specific Terms

(last updated May 14, 2026)

These Service Specific Terms supplement the Inworld AI Terms of Service (currently located at www.inworld.ai/terms) or the Master Services Agreement you signed with Inworld AI (the “Agreement”) and applies to your access and use of our Services, including any Inputs you provide and Outputs you create. It also applies to your use within and outside our Website and our Services, whether directly or indirectly, as well as any attempts to engage in such use. Capitalized but undefined terms used herein have the meanings set forth in the Agreement.

Customer-Hosted Deployment Terms

Preamble

These Customer-Hosted Deployment Terms apply to Customers that deploy the Model(s), and associated Services on Customer’s IT Infrastructure.

Definitions

The capitalized terms in these Deployment Terms will have the meaning set out below:

“Authorized Individuals” means Customer’s employees authorized to access the Model’s source code and weights.
“Customer-Hosted Deployment” means the deployment of the Model by Customer on Customer’s IT Infrastructure.
“Customer’s IT Infrastructure” means the IT infrastructure used by Customer to deploy the Model, including Customer’s private servers or private cloud. Customer’s IT Infrastructure is not provided by Inworld AI and cannot be accessed by Inworld AI unless as expressly authorized by Customer under this Agreement.

All capitalized terms not defined in these Deployment Terms have the meanings set forth in the MSA.

The Model and the Services

2.1. The Model

2.1.1. Technical requirements

The Documentation sets out Inworld AI general recommendations regarding the configuration of Customer’s IT Infrastructure. However, such recommendations are not (a) mandatory or prescriptive, (b) personalized for Customer or Customer’s IT Infrastructure, and (c) comprehensive of every standard infrastructure requirement. Customer is solely responsible for configuring Customer’s IT Infrastructure to support the use of the Model and fixing any issues resulting from Customer’s IT Infrastructure. Customer acknowledges that if Customer does not comply with the Documentation Customer may not be able to use the Model properly and potential issues may occur. Except as otherwise set forth in the applicable Order Form, Inworld AI will not provide any configuration or set-up Service beyond the Documentation.

2.1.2. Delivery of the Model

Inworld AI will deliver the Model and the associated Documentation to Customer on the delivery date set forth in the Order Form, via a mutually agreed transfer method.

2.1.3. Deployment and Fine-Tuning of the Model

Customer is responsible for the Customer-Hosted Deployment of the Model on Customer’s IT Infrastructure in accordance with the Documentation. Except as otherwise set forth in the applicable Order Form, Inworld AI will not provide any deployment or Fine-Tuning services.

2.1.4. Use of the Model

Customer acknowledges that Inworld AI has no control over Customer's use of the Model on Customer’s IT Infrastructure. Consequently, Customer acknowledges that Customer is solely responsible for the use of the Model and for ensuring that such use complies with the terms of the Agreement and all applicable laws and regulations. For the avoidance of doubt, Inworld AI will not be responsible for any of the data loss alteration, or interception for data stored on Customer’s IT Infrastructure or for any interruptions, delays, service failures or other problems inherent with Customer’s IT Infrastructure.

Without limiting the foregoing, Customer is responsible for:

Defining Customer’s Use-Case for the use of the Model;
Configuring Customer’s IT Infrastructure for the purpose of using the Model optimally;
Using the Model only in the context of its business operations and in accordance with this Agreement, and not for illicit, prohibited and/or illegal purposes, to harm or cause damage or injuries to third parties or Inworld AI;
Complying with the Applicable Data Protection Law (including, but not limited to the EU AI ACT) when developing, deploying or using the Model;
Mobilizing an adequate number of qualified employees to ensure the collaboration and effective communication with Inworld AI’s teams and the orderly deployment and management of the Model;
Providing a disclaimer to End-Users of the Model. Such disclaimer should highlight the potential for inaccuracies and unpredictability of the Outputs and encourage End-Users to check important information; and
Ensuring the security of Customer’s IT Infrastructure and Customer Data.

2.2. The Services

Inworld AI will provide Customer's teams with all reasonable guidance and information necessary to facilitate the optimal use of the Model. Any additional support provided by Inworld AI are set forth in the applicable Order Form.

Voice and Speech Services (including text to speech and speech to text)

Voice Content and User Voice Models; Inworld AI Models

(a) Inputs and Outputs. You may transmit or otherwise provide data and information as input to our Services ("Input"). When you provide Input to the Services, you may receive audio output generated and returned by one or more Inworld AI Voice Models (defined below), or text output generated and returned by one or more Inworld AI LLMs (defined below), based on Input (the Output) (Input and Output, collectively, the “Voice Content”). Input may include, without limitation, recordings of your voice, text descriptions, or any other content that you may provide to us through the Services. Your access to and use of the Services, including for the purposes of providing Input to the Services and receiving and using the Output from the Services, is subject to our Acceptable Use Policy. We may enable you to download Output from some (but not all) of the Services; in such cases, you are permitted to use such Output outside of the Services but always subject to these Terms and our Acceptable Use Policy. If you choose to make any of your information publicly available through the Services or otherwise, you do so at your own risk.

(b) User Voice Models. Some of our Services allow you to create a voice model that can be used to generate Output in the form of synthetic audio sounding like your voice or a voice you are authorized to share with us (a "User Voice Model"). To create a User Voice Model through our Services, you may be asked to upload audio recordings of your voice or the voice you are authorized to share with us as Input to our Services, and Inworld AI is permitted to use those audio recordings of the voice you provide to us subject to these terms. For more information on how we collect, use, share, retain, and destroy your audio recordings, please see the below:

Purpose	Description	Legal processing grounds (where required by law)
Collecting your audio recordings and other Voice Data for use in connection with our Services	We collect the audio recordings of your voice that you choose to share with us, as well as any other Voice Data to provide our Services to you.	The processing is necessary for entering into, or performance of a contract to which you are a party
Generating your voice model or otherwise interpreting your voice if you use it for Services such as speech synthesis, conversational AI, dubbing, or translation	We may process the audio recordings of your voice that you provide to us, which include your Personal Data, such as your Voice Data as well as any Personal Data you choose to provide, to provide conversational AI services or to create a voice model that will generate synthetic audio sounding like your voice, as further described below in these terms and the privacy policy.	The processing is necessary for entering into, or performance of a contract to which you are a party.

(i) Except as expressly set forth herein, as between you and Inworld AI, you retain all rights in and to your Input.

(ii) For the avoidance of doubt, Output may be generated by, but does not include, Inworld AI’s foundational and other artificial intelligence voice models (the “Inworld AI Voice Models”) or Inworld AI’s foundational and other artificial intelligence or language learning models (“Inworld LLMs”) (Inworld AI Voice Models and Inworld AI LLMs, collectively, the “Inworld AI Models”). Except as expressly set forth herein, as between you and Inworld AI, you retain all rights in and to your Output.

(d) License to Your Content. You hereby grant to Inworld AI a license to use, reproduce, modify, adapt, publish, translate, create derivative works from, distribute, publicly or otherwise perform and display, and use your Input to provide the Services, to improve the Services, and to ensure the safety and security of the Services. For avoidance of doubt, to the extent that your Input includes your voice, the foregoing license allows Inworld AI to reproduce, modify, publish, create derivative works from, distribute, publicly or otherwise perform, and use your voice, and other indicia of your persona that may be contained therein, to provide the Services.

(e) License to User Voice Models. To the extent you own or acquire any intellectual property rights in or to any User Voice Models, you hereby grant to Inworld AI a license to use, reproduce, modify, adapt, publish, translate, create derivative works from, distribute, publicly or otherwise perform and display, and use the User Voice Models to provide the Services and to improve the Services. .

(f) Necessary Rights. You may not provide Input or create Output for which you do not have all the rights necessary to grant us the license described above. You represent and warrant that the Content and User Voice Models, and our use of the Content and User Voice Models, will not violate any rights of any person or entity, or cause injury to any person or entity.

(g) No PHI. You may not provide any Input that includes protected health information as defined by the Health Insurance Portability and Accountability Act (HIPAA) of 1996, Pub. L. No. 104-191.

Managed GPU Services

Definitions

For purposes of these GPU Terms, the following capitalized terms have the meanings set forth below. Each definition is consistent with, and incorporated from, the MSA; in the event of any inconsistency, the MSA definition controls.

“Acceptable Use Policy” means or “AUP” the then-current acceptable use policy located at inworld.ai/aup/ and as set forth in Section 8 (Acceptable Use Policy) of these GPU Terms.
“Agreement” means collectively, (a) the MSA, (b) all Order Forms, (c) the applicable Service Specific Terms (including these GPU Terms), and (d) the Online Terms.
“Customer Application” means any software application, workload, or service that Customer develops, deploys, or operates using the Managed GPU Services.
“Customer Data” means any of the following: (a) Customer Feedback, (b) Input, and (c) Output, in each case as defined in the MSA.
“Documentation” means the technical documentation made available by Inworld AI describing the features, functionality, and use of the Managed GPU Services, as may be amended from time to time.
“GPU Instance” means a virtual or dedicated GPU computing resource provisioned to Customer through the Managed GPU Services.
“High-Risk Activities” means activities where the use or failure of the Managed GPU Services could lead to death, personal injury, or severe environmental damage, including use in connection with nuclear facilities, weapons systems, manned aircraft or spacecraft navigation, life support systems, or emergency services.
“Managed GPU Services” means the GPU-accelerated cloud infrastructure services provided by Inworld AI, which may include: (a) dedicated or shared GPU Instances; (b) managed inference endpoints; (c) provisioned throughput capacity; (d) associated compute, storage, and networking resources; and (e) management tooling and APIs, each as described in the applicable Order Form and these GPU Terms.
“Order Form” means an executed order form or similar ordering document that references the MSA and specifies the Managed GPU Services to be provided by Inworld AI to Customer and any Fees payable thereunder.
“Restricted Party” means any individual or entity that is (a) listed on, or owned or controlled by a party listed on, any U.S., EU, U.K., U.N., or other applicable sanctions or restricted-party list (including the U.S. Department of the Treasury’s Office of Foreign Assets Control (OFAC) Specially Designated Nationals and Blocked Persons List, the U.S. Department of Commerce Bureau of Industry and Security (BIS) Entity List, Denied Persons List, Unverified List, or Military End User List, and any equivalent list maintained under EU, U.K., or U.N. measures); (b) ordinarily resident in, or organized under the laws of, any country or territory that is the subject of comprehensive U.S., EU, U.K., or U.N. sanctions (including, as of the Effective Date, Cuba, Iran, North Korea, Syria, and the Crimea, so-called Donetsk People’s Republic, and so-called Luhansk People’s Republic regions of Ukraine) (each, a “Sanctioned Jurisdiction”); or (c) acting on behalf of any of the foregoing.
“Service Data” means data generated by the Managed GPU Services in the course of providing the services, including logs, telemetry, and performance metrics, excluding Customer Data.
“Services” means the services provided by Inworld AI to Customer under the Agreement as described in the applicable Order Form, which may include AI model services, Managed GPU Services, and any other services specified therein.

Service Description

2.1 Provision of Services. Inworld AI will provide the Managed GPU Services as described in the applicable Order Form. The Managed GPU Services may include dedicated GPU Instances, managed inference endpoints, and associated infrastructure. Inworld AI will use commercially reasonable efforts to make the Managed GPU Services available consistent with the MSA, these GPU Terms, the Online Terms, and the applicable Order Form.

2.2 Documentation. Inworld AI will make the Documentation available to Customer and may update the Documentation from time to time.

2.3 Modifications to Services. Inworld AI may modify the Managed GPU Services from time to time. If a modification materially reduces the functionality of the Managed GPU Services, Inworld AI will provide Customer with at least thirty (30) days’ prior written notice. If Customer objects to such modification, Customer may terminate the applicable Order Form upon written notice within ten (10) days of receiving notice of the modification.

2.4 No Guaranteed Availability or Capacity. Customer acknowledges that GPU capacity, hardware types, geographic availability, and provisioning timelines may vary based on market conditions, upstream supplier constraints, maintenance, and infrastructure availability. Except as expressly set forth in an applicable Order Form, Inworld AI does not guarantee the availability of any specific GPU model, region, configuration, or uninterrupted capacity.

2.5 Multi-Tenant Environment. Unless expressly stated otherwise in an applicable Order Form, the Managed GPU Services operate in a multi-tenant environment and infrastructure resources may be shared among customers. Customer should not assume dedicated hardware, isolation, or single-tenant operation absent an express written agreement.

2.6 Infrastructure Reallocation. Inworld AI may migrate, rebalance, or reallocate Customer workloads, GPU Instances, or underlying infrastructure (including across cloud providers, regions, hardware SKUs, or clusters) as reasonably necessary for operational, security, capacity-management, or maintenance purposes, provided that such actions do not materially degrade the Managed GPU Services.

Customer Obligations

3.1 Compliance. Customer will use the Managed GPU Services in compliance with: (a) the MSA or Online Terms (www.inworld.ai/terms); (b) these GPU Terms; (c) the applicable Order Form; (d) the Documentation; and (e) all applicable laws and regulations, including those described in Section 3.5 (Export Controls and Sanctions).

3.2 Account Security. Customer is responsible for: (a) maintaining the confidentiality of its credentials and API keys; (b) all activities that occur under its account; and (c) promptly notifying Inworld AI of any unauthorized access or use.

3.3 Customer Applications. Customer is solely responsible for: (a) all Customer Applications deployed on or operated through the Managed GPU Services; (b) the content transmitted, stored, or processed through the Managed GPU Services; and (c) ensuring that its use does not violate any third-party rights.

3.4 Prohibited Uses. Without limiting Section 4.2 of the MSA or Section 3.5 below, Customer will not use the Managed GPU Services: (a) for any High-Risk Activities; (b) to process or store: (i) protected health information regulated under HIPAA absent a separate Business Associate Agreement; (ii) payment card data subject to PCI DSS unless expressly authorized in writing by Inworld AI; or (iii) classified government information or other data requiring safeguarded environments not expressly supported by the Managed GPU Services; (c) to engage in cryptocurrency or blockchain mining unless expressly authorized; (d) to conduct denial-of-service attacks, network abuse, or unauthorized access to any system; or (e) to resell, redistribute, or provide access to the Managed GPU Services to any third party without Inworld AI’s prior written consent.

3.5 Export Controls and Sanctions. Customer acknowledges that the Managed GPU Services, related software, technology, and any deliverables provided under the Agreement (collectively, the “Controlled Items”) are subject to U.S. export control laws and regulations, including the Export Administration Regulations (15 C.F.R. Parts 730–774, the “EAR”), the International Traffic in Arms Regulations (22 C.F.R. Parts 120–130, the “ITAR”), economic sanctions administered by the U.S. Department of the Treasury’s Office of Foreign Assets Control (“OFAC”), and the export control, sanctions, anti-boycott, and customs laws of any other applicable jurisdiction (collectively, “Export Control Laws”). Customer will, and will cause its Affiliates, End-Users, contractors, and any party accessing the Managed GPU Services through Customer to, comply with all Export Control Laws at all times.

Without limiting the foregoing, Customer will not, and will not permit any third party to, directly or indirectly: (a) export, re-export, transfer, release, or provide access to the Managed GPU Services, any Controlled Items, or any output thereof to or for the benefit of (i) any Restricted Party, (ii) any Sanctioned Jurisdiction, or (iii) any national of a Sanctioned Jurisdiction, except as expressly authorized in writing by Inworld AI and by all applicable licenses or authorizations issued by the relevant governmental authority; (b) use the Managed GPU Services to design, develop, produce, or support nuclear, chemical, biological, missile, or other weapons of mass destruction, or to support unsafeguarded or sensitive nuclear fuel-cycle activities, or for any end use prohibited under the EAR (including those in 15 C.F.R. Part 744) or other Export Control Laws; (c) use the Managed GPU Services to train, fine-tune, evaluate, or host any artificial intelligence model on behalf of, or for the benefit of, any Restricted Party; (d) circumvent, attempt to circumvent, or assist any party in circumventing the Export Control Laws, including by re-routing access, falsifying end-user information, or obfuscating end use or end users; or (e) take any action that would cause Inworld AI to be in violation of, or subject to penalties under, the Export Control Laws.

Customer represents and warrants on a continuing basis that: (a) neither Customer nor any of its Affiliates, directors, officers, employees, End-Users, or representatives with access to the Managed GPU Services is a Restricted Party; (b) Customer is not located, organized, or ordinarily resident in any Sanctioned Jurisdiction; (c) Customer has implemented and will maintain reasonable export-control compliance procedures (including end-user screening and end-use diligence) appropriate to its use of the Managed GPU Services; and (d) Customer has obtained, and will maintain, all governmental authorizations, licenses, and consents required for its use of the Managed GPU Services under the Export Control Laws. Customer will promptly notify Inworld AI in writing if Customer becomes aware of any actual or suspected violation of this Section 3.5 or any change in circumstances that would render any of the foregoing representations untrue, and will reasonably cooperate with any audit or investigation conducted by Inworld AI or any governmental authority regarding compliance with the Export Control Laws. Inworld AI may suspend or terminate the Managed GPU Services immediately, without liability, upon any actual or reasonably suspected breach of this Section 3.5. The obligations in this Section 3.5 survive expiration or termination of the Agreement.

3.6 Data Residency. Customer acknowledges that Customer Data and Service Data may be processed and stored in the United States and in other jurisdictions where Inworld AI or its subprocessors maintain operations, unless otherwise expressly agreed in writing in an applicable Order Form or in the DPA (defined below).

3.7 Data Protection; Lawful Basis. To the extent Customer Data includes personal data subject to Applicable Data Protection Law, the parties’ data processing obligations are governed by Inworld AI’s Data Processing Addendum, available at https://inworld.ai/data-processing-addendum (the “DPA”), which is incorporated into the Agreement by reference and, with respect to the processing of personal data, controls over any inconsistent provision of these GPU Terms. Customer is solely responsible for obtaining all rights, consents, notices, and lawful bases necessary for the collection, transfer, and processing of Customer Data through the Managed GPU Services.

Fees and Payment

4.1 Fee Structure. Fees for the Managed GPU Services will be set forth in the applicable Order Form and may include: (a) usage-based fees calculated on compute hours, token consumption, or other metrics; (b) reserved or committed-use fees; (c) data transfer and storage fees; and (d) support fees.

4.2 Metering. Inworld AI’s measurement of Customer’s usage of the Managed GPU Services is final and binding for billing purposes.

4.3 Committed Use. If an Order Form specifies Committed Use, Customer’s payment obligations for the committed capacity are non-cancellable and payable regardless of actual usage. Early termination of a Committed Use order will require payment of all remaining fees through the end of the commitment term.

4.4 Price Changes. Except as otherwise expressly set forth in an applicable Order Form (for example, fixed or committed-use pricing for a stated term), Inworld AI may modify pricing for the Managed GPU Services at any time, and any such change will take effect upon posting or upon written notice to Customer (including by email). Continued use of the Managed GPU Services following the effective date of a price change constitutes Customer’s acceptance of the change. For the avoidance of doubt, prices fixed in an Order Form will remain in effect for the term specified in that Order Form and will not be modified by Inworld AI during that term except by mutual written agreement of the parties.

4.5 Resource Management. Inworld AI may implement rate limits, throughput controls, concurrency restrictions, or other resource-management measures reasonably necessary to preserve the security, stability, or fair allocation of the Managed GPU Services or to address abusive or anomalous workloads. Inworld AI will use reasonable efforts to provide advance notice of any non-emergency throttling that would materially affect Customer’s usage.

Technical Support

5.1 Standard Support. Inworld AI will provide standard technical support for the Managed GPU Services as described in the Documentation, including email-based support during Inworld AI’s standard business hours.

5.2 Enhanced Support. Enhanced support tiers (for example, 24/7 support, dedicated account management, or priority response times) may be available for an additional fee as specified in the applicable Order Form.

5.3 Incident Response. Inworld AI will maintain incident response procedures and will notify Customer promptly of any security incident that materially affects Customer Data, in accordance with the MSA and the Online Terms.

Other Requirements and third party terms

6.1 GPU Hardware. The Managed GPU Services may utilize third-party GPU hardware (for example, NVIDIA or similar). Customer’s use of such hardware through the Managed GPU Services may be subject to additional third-party terms, which Inworld AI will make available to Customer.

6.2 Open-Source Software. The Managed GPU Services may incorporate open-source software components. Applicable open-source licenses will be identified in the Documentation.

6.3 Upstream Dependencies. The Managed GPU Services depend on third-party infrastructure providers, telecommunications providers, and hardware suppliers. Inworld AI is not responsible for delays, failures, interruptions, or degradation of the Managed GPU Services to the extent caused by such third parties or by events outside Inworld AI’s reasonable control.

6.4 Service Data. Inworld AI may collect, generate, and use Service Data for purposes including operating, securing, monitoring, improving, capacity planning, billing, analytics, and developing the Managed GPU Services and Inworld AI’s other products and services, provided that Service Data does not include Customer Data except in de-identified or aggregated form that is not reasonably re-identifiable as relating to Customer or its End-Users.

6.5 Benchmarking. Customer will not publicly disclose benchmarking, performance, capacity, or comparative testing results relating to the Managed GPU Services without Inworld AI’s prior written consent, except as required by law. Internal benchmarking conducted by Customer for its own evaluation and procurement purposes is permitted, provided that the results are kept confidential and are not disclosed to third parties other than under written confidentiality obligations no less protective than the MSA.

Suspension

7.1 Emergency Suspension. Inworld AI may immediately suspend Customer’s access to the Managed GPU Services without prior notice if: (a) required by applicable law or governmental order; (b) Customer’s use poses a security risk to the Managed GPU Services or to other customers; (c) Customer violates the AUP, the Online Terms, or Section 3.5 (Export Controls and Sanctions); or (d) suspension is necessary to prevent material harm to Inworld AI’s infrastructure or third parties. Inworld AI will notify Customer as soon as reasonably practicable. Termination of the Agreement, the Order Form, and the Managed GPU Services is governed by the MSA.

Managed GPU Services — Acceptable Use Policy

This Acceptable Use Policy (the “AUP”) governs Customer’s use of the Managed GPU Services and supplements the MSA, these GPU Terms, and the Online Terms available at www.inworld.ai/terms. Violation of this AUP may result in suspension or termination of the Managed GPU Services in accordance with the MSA.

General Conduct

Customer will not use the Managed GPU Services to: (a) violate any applicable law, regulation, or governmental order; (b) infringe or misappropriate any third party’s intellectual property, privacy, or other rights; (c) transmit, store, or process any content that is unlawful, defamatory, obscene, or that promotes violence or discrimination; (d) engage in or facilitate any fraudulent, deceptive, or misleading activity; (e) distribute malware, viruses, or other harmful code; (f) interfere with, disrupt, or impose an unreasonable burden on the Managed GPU Services or connected networks; (g) circumvent or attempt to circumvent any AI safety controls, content-moderation systems, or usage restrictions imposed by Inworld AI or any third-party model provider whose models are made available through the Managed GPU Services; or (h) train, fine-tune, evaluate, or host any model intended primarily for unlawful mass surveillance, social scoring, or biometric identification prohibited under applicable law.

Security

Customer will not: (a) attempt to gain unauthorized access to any system, network, or account; (b) conduct vulnerability scans, penetration tests, or security assessments against the Managed GPU Services without Inworld AI’s prior written consent; (c) launch or facilitate denial-of-service attacks from or through the Managed GPU Services; (d) circumvent or disable any security features or access controls of the Managed GPU Services; or (e) engage in large-scale credential stuffing, account takeover attacks, automated exploitation, or abusive scraping activity directed at any system, whether or not such system is owned or operated by Inworld AI.

Resource Usage

Customer will not: (a) engage in cryptocurrency or blockchain mining or proof-of-work operations unless expressly authorized in the applicable Order Form; (b) resell, sublicense, or provide access to the Managed GPU Services to any third party without Inworld AI’s prior written consent; or (c) use the Managed GPU Services in a manner designed to circumvent usage limits, billing mechanisms, or other technical restrictions.

Content Restrictions

Customer will not use the Managed GPU Services to generate, store, transmit, or distribute: (a) child sexual abuse material (CSAM) or any content that exploits minors; (b) content designed to facilitate terrorism or violent extremism; (c) non-consensual intimate imagery; or (d) content that facilitates the creation of biological, chemical, nuclear, or radiological weapons.

Enforcement

Inworld AI reserves the right to investigate suspected violations of this AUP and to take any action it deems appropriate, including suspending or terminating access to the Managed GPU Services. Inworld AI may report suspected illegal activity to law enforcement. Customer agrees to cooperate with Inworld AI in investigating any suspected violation. Inworld AI’s failure to enforce any provision of this AUP does not constitute a waiver of its right to do so.

Modifications

Inworld AI may modify this AUP from time to time by posting the updated version, including at www.inworld.ai/terms or inworld.ai/aup/. Continued use of the Managed GPU Services after the effective date of a modification constitutes acceptance of the modified AUP.