Alida's Commitment to Responsible AI Use

Where can I find a list of all the Alida AI features? All new AI features are announced via in-app notification, email, and updated in the complete list of AI features.

Will our company's data be used to train AI models? No. Alida contractually guarantees your data is never used to train the commercial generative AI models we employ. Our primary partner, AWS, also guarantees that customer data submitted to the Amazon Bedrock service is not used to train their original base models.

What core AI technology does Alida use for its generative features? Alida's generative AI (GenAI) capabilities are built on AWS Bedrock, a fully managed service from AWS. We primarily use the industry-leading Anthropic Claude family of models provided through this secure service but continue to evaluate other options for future consideration.

How is our data kept secure during AI processing? We use multiple layers of security. All data is encrypted at rest using AES-256 and in transit using TLS 1.3. Critically, all data exchanged between Alida and AWS Bedrock is secured using AWS PrivateLink, which creates a private connection so your data never traverses the public internet.

Are the AI features optional? Can we turn them off? Yes. Use of Alida's AI capabilities is entirely at your discretion. You have ultimate control via a single, administrative "AI & Machine Learning" toggle within the platform's settings, which allows your organization to enable or disable all AI features at any time.

Who are Alida's main subprocessors for AI? Our primary subprocessor for generative AI is Amazon Web Services (AWS). For specialized capabilities like video transcription, we use carefully vetted partners such as Voxpopme and Discuss.io. Every subprocessor is bound by stringent data protection terms.

How does Alida handle potential bias in AI? We are committed to the ethical use of AI and mitigating bias. Our strategy includes selecting foundation models from partners like Anthropic who demonstrate a strong commitment to fairness, and we continuously evaluate our AI features to ensure they produce equitable and reliable results.

How is Personally Identifiable Information (PII) handled by the AI? Protecting PII is paramount. As a customer, you retain full control to exclude any questions or data fields containing sensitive PII from being included in any AI analysis.

The Strategic Vision & Partnership

A Framework of Trust

Alida's strategic vision for AI is to enhance the efficiency and scalability of your research programs. We have integrated advanced Generative AI (GenAI) into our community-centered research platform to automate manual and administrative tasks. This empowers your researchers to focus on high-impact activities like analyzing data, creating compelling insights, and making better data-driven decisions. Users have the discretion to utilize Alida's optional AI capabilities.

This document outlines our "Trust by Design" philosophy, which is foundational to our product. It details our unwavering commitment to responsible, ethical, and secure AI, ensuring your data privacy and control are never compromised. Alida's strategy is built upon a foundational commitment to responsible and ethical AI use.

The Benefits of GenAI in User Research

By leveraging GenAI, we provide a transformative approach to understanding and leveraging customer insights. This allows your organization to:

Accelerate Speed-to-Insight: Go beyond traditional analysis methods to contextualize data with large language models (LLMs), turning customer feedback into actionable insights faster than ever before.
Enhance Qualitative Analysis: Unearth nuances in patterns and sentiment, even within unstructured text data, achieving human-quality analysis and output from qualitative studies at scale.
Improve Operational Efficiency: Automate manual tasks such as identifying key themes in open-ended responses, translating survey questions, and transcribing video feedback, freeing up researchers to focus on strategic initiatives.
Drive Better Decisions: Gain deeper, more accurate, real-time insights to improve decision-making, enhance customer experiences, and conduct more effective research.

A Partnership for Innovation: Our AI Approval Model

To help you realize the full benefits of AI as quickly as possible, we’ve designed this document to streamline the security and compliance process. We understand that each new AI feature can trigger a new, time-consuming security and privacy review, creating delays that prevent your teams from leveraging powerful new capabilities. This document is designed to make this process more efficient for your organization.

This document provides a single, comprehensive review of the Alida AI Infrastructure. The goal is to support your internal due diligence processes. This document serves as a central, living resource that provides the detailed answers your security, privacy, and legal teams need. By reviewing this foundational framework once, you gain assurance that all future AI features will be held to the same high standards for security, data handling, and governance outlined here. This consistency means you can avoid repetitive reviews for each new capability.

This partnership approach empowers your teams to access new, secure AI-powered features as they become available, ensuring your organization remains at the forefront of research without compromising on security. We are committed to maintaining these standards and will proactively update this document to reflect any relevant changes to our AI infrastructure. We will also continue to update the Alida Subprocessor List as appropriate.

The Alida AI Infrastructure

The Alida AI Infrastructure: Secure, Scalable, and Reliable

When you approve the Alida AI Infrastructure, you are approving a secure, best-in-class technology stack designed for enterprise-grade performance and security.

Core AI Service: Alida's GenAI capabilities are built on Amazon Bedrock, a fully managed service from AWS that provides secure access to a choice of high-performing foundation models, which includes the industry-leading Anthropic Claude family of models. To ensure flexibility and capacity, we also plan to incorporate the AWS Nova family of models, also through Amazon Bedrock.
Secure, Private Connectivity: All data exchanged between Alida's environment and AWS Bedrock is secured using AWS PrivateLink. This creates a private connection that ensures your data never traverses the public internet, dramatically reducing threat vectors and enhancing data security.
High Availability and Performance: To provide our customers with timely access to the latest models and ensure high availability, we leverage AWS's Cross-Region Inference Service (CRIS). This service is explained in detail in the Data Processing and Encryption section.

Alida AI Features & Technology

Alida uses enterprise-grade, secure AI infrastructure for all enabled AI features within the Alida platform.

Alida AI Features & Technology Matrix

Discovery

Optional AI features that support search.

Features	Inputs	Outputs	Sub-Processor	Model(s) Used*	Data Processing Location	Alternative to AI features
Alida Discovery	Natural language queries to find surveys and survey questions	Ranked survey entities; question/attribute data; relevance scores; natural language summaries and navigation links.	Amazon Web Services (AWS)	Anthropic Claude; Amazon Titan Embeddings G1 - Text	Alida (within AWS)	User can search by keyword directly from the Activities page.

Design & Authoring

Optional AI features supporting the design and creation of research studies.

Features	Inputs	Outputs	Sub-Processor	Model(s) Used*	Data Processing Location	Alternative to AI features
AI Survey Creator	Word document files containing drafted survey questions uploaded by clients	A survey generated based on the drafted survey questions in uploaded Word documents	AWS Bedrock	Anthropic Claude	Alida (within AWS)	User can manually create a survey and copy and paste content from a Word document.
Survey translation	The user has the option to input text to the AI in order to design and author research.	The user chooses whether to insert the proposed generated content and the user can edit through the standard editors.	AWS Bedrock	Anthropic Claude	Alida (within AWS)	User can conduct their own translations outside of the Alida platform and copy the translated text into the Alida platform.

Distribution

Optional AI features that allow the user to select who to send the research to.

Features	Inputs	Outputs	Sub-Processor	Model(s) Used*	Data Processing Location	Alternative to AI features
AI Member Group Builder	The user has the option to input text to the AI to define the desired distribution groups.	The AI provides the desired distribution group within Alida.	AWS Bedrock	Anthropic Claude	Alida (within AWS)	User can use the manual Member Group Builder.
AI Email Assistant	The user has the option to input the email text and any additional guidelines on what they would like changed to the AI to ask it to come up with recommended changes to the text.	The AI provides alternative text based on the text and any guidelines provided.	AWS Bedrock	Anthropic Claude	Alida (within AWS)	User edits the text in the Alida platform or copies and pastes it from a third party system into the Alida platform.

Collection

Optional AI features that prompt the participant for information.

Features	Inputs	Outputs	Sub-Processor	Model(s) Used	Data Processing Location	Alternative to AI features
AI Follow-Up	If enabled by the user, the participant provides responses to the AI prompts, using text.	The AI responds to the participant using text, without the user’s direct engagement. The user will be able to review the unprompted and AI prompted content to determine how to use the information for analysis.	AWS Bedrock	Anthropic Claude	Alida (within AWS)	User uses the open-end question types without AI follow-ups.

Analysis

Optional AI features that enable the analysis of quantitative and qualitative results gathered within the Alida platform.

Features	Inputs	Outputs	Sub-Processor	Model(s) Used*	Data Processing Location	Alternative to AI features
Open-end translation	The user has the option to have the AI analyze the text, image, audio, and video responses from the participant as collected by the Alida platform.	The user will be able to review the AI analysis and can review the original source for all data analyzed by the AI to validate accuracy and decide how to use the AI generated data.	AWS Bedrock	Anthropic Claude	Alida (within AWS)	User can view the original responses and export to do translation outside of the Alida platform.
Smart groupings of other/specify and Open End responses	The user has the option to have the AI analyze the text, image, audio, and video responses from the participant as collected by the Alida platform.	The user will be able to review the AI analysis and can review the original source for all data analyzed by the AI to validate accuracy and decide how to use the AI generated data.	AWS Bedrock	Anthropic Claude	Alida (within AWS)	User can view the original responses and export to do further analysis outside of the Alida platform.
AI Image Moderation for Image Upload questions in modern reports	Images uploaded by survey participants through the Image Upload question type, submitted for display in modern reporting.	Images are either displayed (pass) or obfuscated/blocked (fail) in the modern reporting image browser based on content classification. Note: Exported/downloaded images are not moderated; originals are preserved in exports	AWS Amazon Rekognition	AWS Rekognition Image Moderation API — multi-level content classification model (no Anthropic/GenAI model; rule-based AI classifier)	Alida platform (within AWS) Available on all pods except NA1 (ca-central-1), uses NA2 without saving any data to permanent storage	Researchers manually review and download image ZIP exports to inspect and filter inappropriate content themselves. No automated in-platform filtering.
AI Topic and Sentiment Analysis	The user has the option to have the AI analyze the text, image, audio, and video responses from the participant as collected by the Alida platform.	The user will be able to review the AI analysis and can review the original source for all data analyzed by the AI to validate accuracy and decide how to use the AI generated data.	AWS Bedrock	Anthropic Claude	Alida (within AWS)	User can view the original responses and export to do further analysis outside of the Alida platform.
AI Assistant for reporting	The user has the option to have the AI analyze the text, image, audio, and video responses from the participant as collected by the Alida platform.	The user will be able to review the AI analysis and can review the original source for all data analyzed by the AI to validate accuracy and decide how to use the AI generated data.	AWS Bedrock	Anthropic Claude	Alida (within AWS)	User can use all the reporting features without the AI Assistant, and can export to develop their reports outside of the Alida platform.
Video Analysis	The user has the option to have the AI analyze the text, image, audio, and video responses from the participant as collected by the Alida platform.	The user will be able to review the AI analysis and can review the original source for all data analyzed by the AI to validate accuracy and decide how to use the AI generated data.	Twelve Labs	Pegasus	Alida (within AWS)	User can view the video.

Community Management

Optional AI features that enable engagement of the community.

Features	Inputs	Outputs	Sub-Processor	Model(s) Used*	Data Processing Location	Alternative to AI features
AI Copywriter for Hub Posts	The user has the option to input text to the AI to generate content for the community.	User chooses whether to insert the proposed generated content and the user can edit through the standard editors.	AWS Bedrock	Anthropic Claude	Alida (within AWS)	User edits the text in the Alida platform or copies and pastes it from a third party system into the Alida platform.

Community Analysis

Optional AI features that enable analysis of the community.

Features	Inputs	Outputs	Sub-Processor	Model(s) Used*	Data Processing Location	Alternative to AI features
Hubs topic and sentiment analysis (available to Alida Professional Services, not directly for customers)	The user has the option to have the AI analyze the text, image, audio, and video responses from the community members as collected by the Alida platform.	The user will be able to review the AI analysis and can review the original source for all data analyzed by the AI to validate accuracy and decide how to use the AI generated data.	AWS Bedrock	Anthropic Claude	Alida (within AWS)	Not applicable as it is not available to user directly.

Video Feedback

Optional transcription, sentiment, and theming.

Features	Inputs	Outputs	Sub-Processor	Model(s) Used*	Data Processing Location	Alternative to AI features
Video Analysis/ Transcription	Video that the participant has agreed to have created	Text analysis of video	Voxpopme	Proprietary (based on IBM Watson)	Voxpopme	User can view the recording and export for further analysis outside of the Alida platform.

Video Discussions

Optional transcription and sentiment.

Features	Inputs	Outputs	Sub-Processor	Model(s) Used*	Data Processing Location	Alternative to AI features
Video Analysis/ Transcription	Video that the participant has agreed to have created	Text analysis of video	Discuss.io	OpenAI Whisper, Deepgram Nova-2, AWS Comprehend	Discuss.io	User can view the recording and export for further analysis outside of the Alida platform.

Legacy Text Analysis

This model is being phased out and is no longer trained.

Optional topic/sentiment analysis (English only).

Features	Inputs	Outputs	Sub-Processor	Model(s) Used*	Data Processing Location	Alternative to AI features
GenAI Text Analysis	The user has the option to have the AI analyze the open-end response from participants captured by the Alida platform.	The user will be able to review the AI analysis and can review the original source for all open-end data analyzed by the AI to validate accuracy and decide how to use the AI generated data.	Alida Custom Model (2019)	Alida Custom Model (2019)	Alida (within AWS)	Not applicable, as no longer offered.

*The specific model families used are subject to change.

Alida Subprocessors

Every Alida subprocessor is carefully vetted and put through a rigorous security, privacy, and compliance review prior to onboarding. Each is bound by contract terms that are as stringent and protective as those set out in our standard Data Protection Schedule, available at https://www.alida.com/security. Our primary subprocessor for GenAI is Amazon Web Services (AWS). All Alida sub-processors are listed here.

Data Usage Policy

While Alida does use anonymized Platform usage data to improve and optimize our products and services, we never use our Customer’s personal or confidential data to train any AI models, and we require the same from all of our partners and subprocessors.

As we move forward, Alida remains committed to providing our customers with a secure, innovative, and ethically-driven AI experience. We appreciate the trust you place in our platform and look forward to continually exceeding your expectations.

Trust & Control: Our Commitment to Data Security, Governance, and Privacy

This section provides a comprehensive overview of the security, privacy, and governance framework that underpins the Alida AI Infrastructure.

Data Security & Privacy

Model Training

Alida contractually guarantees that no customer data is used to train the commercial GenAI models it employs. This commitment is reinforced by our partners. As stated by AWS regarding its Bedrock service: "AWS guarantees that customer data submitted to the Bedrock service is not used to train the original base model."

Data Processing and Encryption

All customer data processed by our AI features is handled within a secure and isolated environment.

Encryption in Transit: Data is encrypted in transit between Alida and AWS Bedrock using strong protocols like TLS 1.3, traversing the secure AWS global network backbone.
Encryption at Rest: All customer data within the Alida platform is encrypted at rest using industry-standard AES-256 encryption.

PII and Sensitive Data Handling

Protecting personally identifiable information (PII) and sensitive data is paramount. Alida provides multiple layers of protection:

Customer Control: You retain full control to exclude any questions containing sensitive data from AI analysis.

Data Residency & Cross-Region Processing for Canadian Customers

Alida is committed to meeting the data residency requirements of our global customers. Our architecture is designed to provide cutting-edge AI capabilities while respecting data sovereignty.

Amazon Bedrock Cross-Region Inference Service (CRIS)

To provide customers with accelerated access to the newest and most powerful AI models, Alida may use AWS's CRIS.

Data Residency and Logging Guarantee

Even when CRIS is used for transient processing, our commitment to data residency remains firm:

All persistent data remains in the customer's source region.
All logs related to AI processing are stored exclusively in the customer's source region.
No inference data, customer content, or logs are ever stored or persisted in the target (inference) regions.

For Canadian Customers

If a specific model or sufficient capacity is unavailable in the Canada (Central) region, CRIS intelligently and securely routes the inference request to another AWS region (e.g., in the US) for processing. The Canadian Centre for Cyber Security (CCCS), for instance, has assessed Amazon Bedrock as compliant for government workloads even when inference occurs in US regions, precisely because the data is processed transiently and is not stored there.

Governance, Ethics, and Customer Control

Mitigating Bias and Ensuring Fairness

Alida is committed to the ethical use of AI. We select foundation models from partners who demonstrate a strong commitment to fairness and bias mitigation. We continuously evaluate our AI features to ensure they produce equitable and reliable results.

Your Control Over AI Features

Ultimate control remains in your hands. Access to all of Alida's AI features is governed by a single, administrative "AI & Machine Learning" toggle within the platform's settings. This allows your organization to enable or disable all AI capabilities at any time.

Alida's Legacy AI Model

Prior to the widespread availability of modern GenAI, Alida developed a custom, English-only machine learning model to power Alida’s legacy Text Analysis feature. This model was trained in 2019 using data exclusively from customers who had explicitly opted in to participate in the training process.

This legacy model is no longer being trained or updated, and no new customer data is being used with it. It is being actively phased out and replaced by our new, superior features powered by AWS Bedrock. This legacy feature is not available in our AP3 (Australia) data center.