AI Features Addendum

Last updated on June 11, 2026.

This AI Features Addendum sets out the specific terms that apply to the use of artificial intelligence and machine learning features on Poocho's Platform. It applies in addition to our Platform Terms and the relevant Product Terms, and forms part of the agreement between you and Poocho Research Technologies Private Limited.

If you have signed a scope of work, master services agreement, or order form with us, that agreement also applies and prevails over this Addendum where there is a conflict. Where a term is defined in the Platform Terms or the relevant Product Terms and used here, it has the same meaning.

By accessing or using the AI Features on the Platform, you agree to this Addendum.

1. Scope of this Addendum

This Addendum applies to the AI Features in Studio, our analysis product. The general terms that apply to Studio — including ownership of Customer Data, our role as a data processor, sub-processors, and your obligations as a Customer — are set out in the Studio Product Terms. This Addendum sits alongside the Studio Product Terms and covers the AI Features specifically.

We may extend this Addendum to cover AI Features in other products on the Platform — including Recruit — in future versions. When we do, we will update this document and give you reasonable notice in line with the change-of-terms process in the Platform Terms.

Where this Addendum is silent on a question, the Studio Product Terms and the Platform Terms apply.

2. Definitions

In this Addendum:

• AI Features means the artificial intelligence and machine learning features available on the Platform, as described in clause 3.
• AI Output means any text, summary, theme, framework, transcript, translation, report, or other content generated by an AI Feature in response to Customer Data or your instructions.
• AI Sub-processor means a third-party service provider listed on the Sub-processor List that provides an artificial intelligence or machine learning capability used in an AI Feature.
• Service Telemetry means operational metadata about how the AI Features are used — for example, request volumes, error rates, latency, and feature engagement — that does not include Customer Data or any personal data of identified or identifiable individuals.

Terms not defined here have the meaning given in the Studio Product Terms or the Platform Terms.

3. The AI Features

The AI Features currently include:

• Transcription — converting audio and video recordings into text transcripts.
• Translation — translating transcripts and text content between supported languages.
• Synthesis — generating summaries, themes, reports, and other analytical outputs across Customer Data.
• Framework analysis — organising Customer Data against an analytical framework you select or define.
• Querying — answering questions you ask about Customer Data in natural language.

We may add, change, or discontinue AI Features from time to time. We will give you reasonable notice of any material change.

To provide the AI Features, we use a combination of our own technology and the AI Sub-processors listed on the Sub-processor List. We keep that list current, and we add or remove AI Sub-processors as our technology stack evolves.

4. How Customer Data flows to AI Sub-processors

4.1 What is sent

When you use an AI Feature, the Customer Data needed for that Feature to run is sent to the relevant AI Sub-processor for processing. The data sent to each Sub-processor depends on its role in the pipeline:

• Transcription and diarisation (Sarvam — primary): the audio file extracted from your recording, converted to 16kHz mono MP3 within Poocho's AWS infrastructure before transmission.
• Transcription (AssemblyAI — fallback): the audio file, accessed by AssemblyAI via a time-limited presigned URL from Poocho's AWS storage, in the event that primary transcription is unavailable.
• Content extraction, summary, and analysis (Google Gemini): transcript text — including chapters, summary, key takeaways, and demographic markers extracted from the transcript. Where translation is required in the fallback transcription path, text segments are also sent to Gemini for translation.
• Semantic search (Cohere): text segments from transcripts and documents, for the purpose of generating vector embeddings only. Cohere does not retain embedding vectors after the API response is returned — they are stored in Poocho's own vector database.
• Chat agent and querying (Anthropic Claude): text — the content of your query and the relevant transcript or document context provided to the agent.

No video frames, no raw video files, and no personally identifiable metadata beyond what is necessary for the Feature in use are transmitted to any Sub-processor. Video files never leave Poocho's AWS infrastructure — audio is extracted from video within AWS before any transmission occurs.

The specific AI Sub-processors, the categories of Customer Data sent to each, the purpose of the transmission, and the retention applied by each provider are set out on the Sub-processor List.

4.2 What is not sent

We do not send to AI Sub-processors any Customer Data that is not required for the AI Feature you are running. We do not send Customer Data to AI Sub-processors for purposes other than providing the AI Feature to you.

4.3 What AI Sub-processors do with Customer Data

Each AI Sub-processor we use processes Customer Data only as needed to provide the relevant capability back to us, and is engaged under service, API, or enterprise terms that include obligations on confidentiality, security, and purpose limitation. Each AI Sub-processor is also subject to the no-training commitment in clause 5, on the basis described on the Sub-processor List.

5. No training on Customer Data

5.1 Our commitment

We do not use Customer Data to train, fine-tune, retrain, or otherwise improve our own artificial intelligence or machine learning models. Customer Data is processed only to provide the AI Features to you and is not used to develop the underlying models that power those Features.

5.2 AI Sub-processor commitments

Each AI Sub-processor that processes Customer Data on our behalf is contractually prohibited from, or has confirmed in writing that it does not, use Customer Data to train, fine-tune, retrain, or otherwise improve its own artificial intelligence or machine learning models, or any models it makes available to other customers. The basis for this commitment for each Sub-processor is set out on the Sub-processor List.

5.3 Limited exceptions

The commitments in clauses 5.1 and 5.2 apply to Customer Data. They do not apply to Service Telemetry, which we and our AI Sub-processors may use to:

• monitor the operation, performance, security, and reliability of the AI Features;
• detect and prevent fraud, abuse, and misuse of the AI Features;
• improve the operational performance of the AI Features (for example, latency, error handling, and capacity planning); and
• generate aggregated and deidentified analytics about how the AI Features are used.

Service Telemetry is operational metadata only. It does not include Customer Data, the content of your transcripts, documents, or queries, or any personal data of identified or identifiable individuals. We do not use Service Telemetry to reconstruct Customer Data or to develop models that operate on the substance of Customer Data.

6. AI Output

6.1 What AI Output is

AI Output is generated by artificial intelligence and machine learning models running over the Customer Data you provide and the instructions you give. It is an aid to your work — not a substitute for your own judgment.

6.2 What AI Output is not

AI Output:

• is not a statement of fact by Poocho about any individual, organisation, or subject matter;
• is not a professional opinion of any kind, including legal, medical, financial, or research advice; and
• is not verified by Poocho for accuracy, completeness, or fitness for any particular purpose.

6.3 Limits of AI

Artificial intelligence and machine learning models can produce output that is inaccurate, incomplete, inconsistent, biased, or otherwise unsuitable for your purposes. This includes producing content that appears confident or detailed but is wrong in substance — sometimes described as "hallucination." This is a known limit of the current technology and is not a defect in the AI Features.

You should review and validate AI Output before relying on it, citing it, sharing it, or acting on it. The level of review appropriate to a given AI Output depends on the use you are putting it to and the sensitivity of the subject matter — for example, AI Output used in a finished deliverable, a regulated context, or a decision that affects an identified individual warrants more careful review than AI Output used as an exploratory aid.

6.4 Your responsibility for use of AI Output

You are responsible for how you use AI Output, including:

• whether to rely on it, share it, or include it in any deliverable;
• the accuracy and completeness of any deliverable that includes or is informed by AI Output; and
• compliance with any laws, regulations, professional standards, or contractual obligations that apply to your use of AI Output, including in regulated sectors.

6.5 Ownership of AI Output

AI Output is Customer Data and is treated in line with clause 3 of the Studio Product Terms. As between you and Poocho, you own AI Output generated from your Customer Data, and our role as data processor for Customer Data extends to AI Output.

7. Acceptable use of AI Features

In addition to the general acceptable-use rules in the Platform Terms, the Studio Product Terms, and the Fair Use and Acceptable Use Policy, you must not use the AI Features to:

• generate content that misrepresents AI Output as a verified statement of fact, a professional opinion, or content authored by a named person without their consent;
• attempt to reverse engineer, extract, or reconstruct any AI Sub-processor's model, training data, or underlying technology through the AI Features; or
• use AI Output to train, fine-tune, or otherwise develop any artificial intelligence or machine learning system that competes with the Platform.

The Fair Use and Acceptable Use Policy sets out additional rules on consumption conduct and how you may use Studio outputs. Those rules apply in addition to this clause.

We may suspend or restrict your access to specific AI Features, or to the Platform as a whole, in line with clause 11 of the Platform Terms if we reasonably believe you are in breach of this clause or the Fair Use and Acceptable Use Policy.

8. Special category and sensitive data in AI Features

The rule in clause 3.5 of the Studio Product Terms — that you should not upload personal data of a special or sensitive nature except where Studio is expressly intended for that purpose and you have a lawful basis — applies in full to the AI Features. Running an AI Feature over Customer Data that contains special category or sensitive data is a use of Studio, and the same restrictions and the same requirement to write to us in advance apply.

9. Changes to AI Sub-processors

We may add, change, or remove AI Sub-processors from time to time as our technology stack evolves. We will update the Sub-processor List to reflect any change.

For material changes — for example, adding a new AI Sub-processor that processes a new category of Customer Data, or moving an AI Feature to a different AI Sub-processor — we will give you reasonable notice through the Platform, by email to the address associated with your account, or by an update to the Sub-processor List, in line with the change-of-terms process in the Platform Terms.

10. Future versions of this Addendum

This Addendum will be updated as the AI Features on the Platform evolve. In particular:

• when AI Features are introduced in products other than Studio, this Addendum will be extended to cover those Features;
• when the technology stack underlying the existing AI Features changes materially, this Addendum will be updated to reflect the new arrangement.

We will give you reasonable notice of any material change to this Addendum in line with the change-of-terms process in the Platform Terms.

11. Other terms

The remaining terms of the Platform Terms and the Studio Product Terms — including governing law, jurisdiction, class action waiver, liability, indemnity, communications, and changes — apply to your use of the AI Features. Where applicable, the terms of any DPA between you and Poocho also apply.

12. Contact

For questions about the AI Features, write to studiodesk@poocho.co.

For DPDPA grievances, contact our grievance officer as set out in clause 22 of the Platform Terms.