PhindorBack to phindor.com
Legal

Terms of Service

These terms govern Phindor’s broader AI infrastructure—from model development and hosted inference to voice, service access, domain intelligence, APIs and research environments.

Phindor LimitedNairobi, KenyaUpdated 27 October 2026
How this document works

This document applies across Phindor’s AI infrastructure where it is presented or incorporated by reference. Product-, research- or enterprise-specific agreements may add to or replace parts of it for a particular engagement.

1. Acceptance and scope

These Terms of Service govern access to and use of Phindor websites, developer tools, APIs, model services, voice and speech systems, hosted infrastructure, AI products, research or pilot environments, JuaFlow and other services where these terms are presented or incorporated by reference (collectively, the “Services”).

By accessing or using a Service, you agree to these terms. If you use a Service for an organisation, you represent that you have authority to bind that organisation. Signed order forms, statements of work, service-level agreements, data-processing agreements, research agreements or other contracts may add to or replace parts of these terms for a particular engagement.

Mandatory rights that cannot lawfully be excluded remain unaffected.

2. Phindor and what we build

“Phindor”, “we”, “us” or “our” refers to Phindor Limited, Kenya. We build AI infrastructure designed to make intelligent systems useful in the environments where people and organisations actually operate.

Depending on the Service, our work may include:

  • localised intelligence: language models, model adaptation, fine-tuning, evaluation, datasets, benchmarks and systems designed for local languages, accents, code-switching, domain terminology and context;
  • voice and speech infrastructure: speech recognition, speech synthesis, transcription, voice agents, calling infrastructure and related APIs;
  • service-access infrastructure: systems that connect AI to knowledge, software, databases, people and real-world actions, including JuaFlow;
  • domain intelligence: models and AI systems designed for particular industries or use cases, including healthcare, manufacturing, agriculture, financial services, public services and workforce operations;
  • model infrastructure: model hosting, inference, routing, APIs, registries, repositories, evaluation environments and compute used to run or test AI models; and
  • research and pilots: experimental systems, prototypes and collaborative programmes used to develop, validate or evaluate AI capabilities.

Not every capability is available in every product, market or plan. Some Services may be experimental, invitation-only, geographically restricted or governed by additional terms.

3. Eligibility, accounts and access

You must provide accurate and current information and keep account credentials, API keys and access tokens secure. You are responsible for activity performed through your accounts, workspaces, integrations and authorised users.

Some Services, models, datasets or downloads may require approval, identity or organisation verification, a stated intended use, geographic eligibility or acceptance of additional licence terms. We may deny or revoke access where necessary to protect security, comply with law or licence conditions, or enforce legitimate access restrictions.

4. Customer and user responsibilities

You are responsible for how you configure and use the Services, including:

  • ensuring you have the legal rights and permissions required for data, datasets, models, model weights, documents, recordings, prompts, APIs and systems you provide or connect;
  • providing required privacy notices and obtaining consent or another lawful basis where applicable;
  • respecting licences, intellectual-property rights, confidentiality obligations and restrictions attached to third-party models, datasets and software;
  • configuring access, approval and human-review controls appropriate to the use case;
  • reviewing outputs, model behaviour and actions where an error could materially affect a person or organisation; and
  • complying with laws, sector rules, communications rules and platform requirements applicable to your use.

5. Customer materials, datasets and models

“Customer Materials” means content or assets you provide, upload, connect or make available to a Service, including documents, messages, audio, transcripts, datasets, labels, prompts, evaluation sets, model weights, adapters, checkpoints, code, records, instructions and connected-system data.

As between you and Phindor, you retain your rights in Customer Materials, subject to the rights of third parties. You grant Phindor and authorised service providers a limited right to host, process, copy, transform, transmit, evaluate and otherwise use Customer Materials only as reasonably necessary to provide, secure, support and administer the relevant Service, follow your lawful instructions, and meet legal obligations.

You represent that you have the rights necessary to provide Customer Materials and to authorise the processing you request. Public availability does not, by itself, mean that content is free of copyright, privacy, database, confidentiality or licence restrictions.

6. Model development, adaptation and evaluation

Where we provide model-development services, we may train, fine-tune, adapt, quantise, convert, benchmark or evaluate models using datasets and instructions defined by the relevant project or agreement.

The ownership and permitted use of a resulting model, adapter, checkpoint, dataset, evaluation set or other deliverable will be determined by the applicable order form, statement of work, licence or project agreement. Unless expressly agreed otherwise, each party retains ownership of technology, data, models and materials it owned before the engagement.

Model-development results depend on training data, base-model licences, compute, evaluation methods and technical constraints. We do not guarantee that a model will achieve a specific level of accuracy, safety, latency, cost or performance unless a written agreement states a measurable commitment.

Private Customer Materials are not used to train general-purpose models for other customers unless you expressly agree to that use in writing.

7. Hosted models, inference and compute

Phindor may host, route or serve AI models through APIs, applications or managed infrastructure. Hosted models may be owned by Phindor, by a customer, by a third-party model developer or made available under an open-source or other licence.

You are responsible for ensuring that any model you ask us to host or serve can lawfully be used for the intended purpose and environment. We may apply technical controls such as rate limits, geographic restrictions, authentication, logging, content safeguards, model routing, quantisation or resource limits.

Inference capacity, latency and availability can vary with model size, traffic, hardware availability, third-party infrastructure and other operational conditions. Any committed service levels apply only where expressly stated in a signed agreement.

8. Model registries, repositories and downloads

Where we provide a model catalogue, registry, repository or download service, model cards, benchmark results, descriptions and metadata are provided to help users evaluate available resources. They are not a guarantee that a model is suitable for a particular use.

Models, datasets and related files may carry separate licences or restrictions imposed by their owners. Those terms remain applicable even when access is provided through Phindor. We may require access requests, intended-use information, organisation details or additional approvals before enabling a download or API.

9. AI outputs and evaluation results

Machine-learning and generative-AI systems are probabilistic. Outputs, transcriptions, classifications, translations, scores, recommendations and benchmark results may be incomplete, inaccurate or sensitive to the input, model version, evaluation method or environment.

You must use appropriate judgment and human review where an error could materially affect safety, health, employment, financial rights, legal rights, access to essential services or other significant interests. Unless we expressly agree otherwise, AI outputs are not a substitute for professional advice or a guarantee of factual correctness.

10. Service access, agents and connected actions

Some Phindor Services, including JuaFlow, connect AI to customer channels, knowledge sources, APIs, databases, enterprise systems and people so a request can be carried through to an outcome.

Where a Service can create, retrieve, update or send information through a connected system, you authorise actions that fall within the permissions and instructions you configure. You are responsible for deciding which actions may run automatically and which require human approval.

We may provide approval gates, validation, permissions, audit trails, rate limits or other safeguards, but those controls do not replace your responsibility to configure connected systems appropriately.

11. Voice, speech and communications

Voice Services may process live or recorded audio, call metadata, transcripts, text-to-speech output, speech features or related signals. If you use calling, recording, transcription, messaging or outreach functionality, you are responsible for providing notices and obtaining consents required by applicable law.

You must not use the Services for unlawful spam, deceptive communications, impersonation, harassment, unauthorised recording or communications prohibited by the relevant telecommunications or platform rules.

12. Domain-specific and higher-risk systems

AI systems used in healthcare, finance, employment, education, government, public services, credit, identity, safety-critical operations or other regulated contexts may require additional controls, validation, contracts, professional oversight, impact assessments or sector-specific compliance.

Unless expressly agreed otherwise, Phindor provides technology infrastructure and does not itself provide medical, legal, financial or other regulated professional advice. We may require additional review or safeguards before supporting a higher-risk deployment.

13. Third-party models, data, software and platforms

The Services may interoperate with third-party model providers, cloud infrastructure, open-source software, datasets, messaging platforms, enterprise software, telephony providers, databases and custom APIs.

Third-party and open-source components remain subject to their own licences, terms, policies, usage restrictions and availability. You must comply with those terms where they apply to your use. Phindor is not responsible for a failure caused solely by a third-party service outside our reasonable control.

14. Acceptable use

You may not use a Phindor Service to:

  • violate law, regulation, court order or the rights of another person;
  • obtain or process data, datasets, recordings or model artefacts without the required rights or authority;
  • gain unauthorised access to accounts, models, systems or networks;
  • introduce malware or intentionally disrupt, overload or abuse infrastructure;
  • facilitate fraud, phishing, identity theft, deceptive impersonation, unlawful surveillance or prohibited discrimination;
  • circumvent geographic, security, licence, usage, billing or access restrictions;
  • use a model or dataset contrary to its applicable licence or owner restrictions; or
  • reverse engineer, scrape, resell or copy protected parts of a Service except where law or a written agreement expressly permits it.

15. Fees, usage and taxes

Paid Services are charged according to the applicable order form, subscription, proposal, usage schedule or agreement. Usage may be measured by compute time, model tokens, inference calls, API requests, storage, model hosting, conversations, calls, speech minutes, actions, downloads, credits or another stated unit.

Unless agreed otherwise, fees exclude applicable taxes, duties and levies. We may suspend or limit paid Services for overdue undisputed amounts after any applicable notice period.

16. Confidentiality

Each party may receive non-public business, technical, research or commercial information from the other. A receiving party will use confidential information only for the relationship, protect it with reasonable care and disclose it only to people or providers who need it and are subject to appropriate confidentiality duties, except where disclosure is required by law.

17. Intellectual property and ownership

Phindor and its licensors own the Services and Phindor technology, software, infrastructure, interfaces, designs, documentation, trademarks, Phindor-owned models, evaluation tooling and other original materials, except for Customer Materials and third-party materials.

Customer ownership of customer-provided models, datasets and other Customer Materials is not transferred to Phindor. Ownership of custom-trained or jointly developed deliverables is governed by the applicable project agreement.

Nothing in these terms grants rights to third-party models, datasets or software beyond the rights provided by their applicable licences.

18. Feedback

If you voluntarily provide product, research or technical feedback, you permit us to use that feedback without restriction or payment. We will not publicly identify you as the source without permission.

19. Security

We use technical and organisational safeguards designed to protect the Services and data we process. Depending on the Service, this may include access controls, authentication, network protections, logging, environment separation, secrets management, monitoring, backup and incident-response procedures.

You must also protect credentials, model access tokens, API keys and administrator access. Report suspected vulnerabilities or security incidents to info@phindor.com.

20. Privacy and data protection

Our Privacy Policy explains how we handle personal data across Phindor websites, model services, research environments, voice systems, JuaFlow and other Services.

Where Phindor processes personal data on behalf of a customer, a data-processing agreement or other contract may add obligations relating to security, subprocessors, retention, transfers, data-subject requests and breach response.

21. Availability, maintenance and service levels

No online or AI service is uninterrupted. Maintenance, model updates, capacity constraints, hardware failures, provider outages, internet failures, customer-system failures, security events and events outside reasonable control can affect availability or performance.

Any uptime, support, recovery, latency or service-credit commitment applies only where stated in a signed agreement.

22. Changes to Services

We may change models, infrastructure, interfaces, providers, features or technical architecture to improve the Services, security, performance or compliance. We will use reasonable efforts to avoid materially reducing contracted functionality during a committed term unless required for security, law or third-party changes.

23. Suspension and termination

We may suspend or restrict access where reasonably necessary to address security risk, unlawful use, licence violations, non-payment, abuse, material breach or a legal requirement. We will provide notice where reasonably practicable.

On termination, data export, deletion, model handover and continued access are governed by the applicable plan or agreement.

24. Disclaimers

Except for commitments expressly stated in a signed agreement and to the extent permitted by law, the Services are provided on an “as available” basis. We do not warrant that every model, output, dataset, benchmark, integration or Service will be error-free, uninterrupted, suitable for every purpose or compatible with every system.

25. Limitation of liability

To the maximum extent permitted by applicable law, neither party will be liable for indirect, incidental, special, exemplary, punitive or consequential losses arising from these terms, including loss of profits, goodwill or data, except where such liability cannot lawfully be excluded.

Any additional liability limits, exclusions or negotiated caps in a signed agreement prevail for that engagement.

26. Indemnity

To the extent permitted by law, you will defend and indemnify Phindor against third-party claims arising from Customer Materials, models, datasets, instructions, connected systems or your unlawful use of the Services where the claim results from your breach of these terms or violation of another person’s rights.

27. Governing law, local law and disputes

These terms are governed by the laws of Kenya unless a signed agreement specifies otherwise. Courts of competent jurisdiction in Kenya have jurisdiction subject to any agreed arbitration or dispute-resolution clause.

Where a Service is deployed in another country, mandatory local laws may also apply. Nothing in these terms removes consumer or data-protection rights that cannot lawfully be waived.

28. Events outside reasonable control

Neither party is responsible for delay or failure caused by events outside its reasonable control, including major internet or cloud failures, power failures, natural disasters, civil disruption, government action, widespread telecommunications failures or similar events, provided the affected party takes reasonable steps to mitigate the impact.

29. Changes to these terms

We may update these terms as our Services, legal obligations or operating model evolve. The current version will be published at this URL. Where a material change affects an existing paid commitment, any notice requirements in the applicable contract will apply.

30. Contact

Questions about these terms can be sent to info@phindor.com.

Phindor Limited · Nairobi, Kenya · phindor.com