AI Policy
Last Updated: October 1, 2025
Videotape Inc.
Last Updated: 9/26/2025
At Videotape, AI amplifies and accelerates human insights – it doesn't replace them. We are committed to developing and deploying this technology in a manner that is secure, fair, and transparent. This Artificial Intelligence Policy explains how we use AI, our commitments to you, and the safeguards we have in place.
WHY DO WE USE AI AT ALL?
Our Approach to AI in Research
At Videotape, AI is not a feature; it's the foundation. Unlike other research companies that may use AI as a chatbot or other direct customer interface, Videotape was built on the principle of deep AI integration. We see its true value as a background process enabler that makes a new scale of research possible.
For us, this means:
It's now possible to conduct distributed, on-camera research at scale for the very first time.
We can parse hours of video and thousands of data points in hours instead of weeks.
We deliver insights that distill results into highly digestible reports anyone can understand.
We've made the tedious process of data exploration and dashboard navigation a thing of the past.
Our Perspective on Synthetic vs. Authentic Research
Some research companies have begun using synthetic AI audiences, or personas, as a stand-in for human audiences. While useful for brainstorming, this isn't real research. The risk of relying on synthetic audiences is that the models are overdetermined by their inputs and tend to confirm the biases of the researcher.
Our approach is the opposite: video interviews at scale to discover and quantify the full range of authentic human responses.
Our Core Principles
Human-centric by design. Our AI is built to augment human understanding, not replace it. It serves as a powerful tool for researchers, automating repetitive tasks and highlighting key moments to make the research process faster and more accessible.
Customer data is not our product. Without express permission, we would never use customer survey data to train AI models. Our internal models are trained on our own data, collected from specific test studies conducted for that purpose.
Privacy and security are paramount. We integrate privacy and security into every stage of our AI development and deployment. We use industry-standard encryption, strict access controls, and data minimization techniques to protect your information.
The Benefits Our Use of AI Has for You
Quant and qual combined in a single instrument. We’ve pioneered a fusion of two complementary methods, delivering statistical robustness and anecdotal texture in a single experience. Get the complete picture with one effort, not two.
Accelerated learning with less effort. AI-powered sentiment analysis, thematic coding, and insight synthesis help you quickly identify key patterns, understand the broader context, and get to the core insights of your research with less effort than ever before.
Ensuring data integrity and safety. We use AI for automated fraud detection to identify inauthentic participants and for content safety moderation to flag inappropriate content. This ensures the feedback you receive is authentic, relevant, and safe for review.
WHAT TYPES OF DATA DO WE PROCESS WITH AI?
We apply AI processing to the following data types:
Video recordings. Participant responses and screen recordings.
Audio data. Voice responses and verbal feedback.
Images. Screenshots and video frames.
Text data. Survey responses, transcripts, open-ended answers.
Tabular data. Structured responses, ratings, demographic information.
Metadata. Timestamps, interaction patterns, engagement metrics.
WHAT DO WE USE AI TO ACCOMPLISH?
We use AI for core research functions, including the following:
Transcription and transcoding. Preparing standardized artifacts for analysis.
Translation. Making insights accessible across dozens of language barriers.
Theme extraction. Identifying patterns and recurring topics.
Sentiment analysis. Understanding the reasons behind the response.
Quality scoring. Ensuring response validity and completeness.
Insight generation. Creating actionable summaries and key findings.
Fraud detection. Identifying bot responses and bad actors.
Content safety. Moderating inappropriate or harmful content.
Risk analysis. Assessing data quality and study validity.
Participant matching. Connecting the right participants to studies.
Anomaly detection. Flagging unusual patterns requiring review.
WHERE DO OUR AI MODELS LIVE?
Videotape’s models operate in two environments:
Internal models. Built and maintained within Videotape’s secure cloud infrastructure on GPU nodes, using a mix of open-source and proprietary models.
Trusted third-party vendors. Including OpenAI, Anthropic, and Google, for specialized tasks such as translation and moderation. All vendors must meet Videotape’s data protection standards and are bound by written agreements prohibiting data reuse. Vendor compliance is reviewed periodically.
IS OUR CUSTOMER SURVEY DATA USED TO TRAIN VIDEOTAPE'S MODELS?
No.
Without express written permission, customer data and results are never used to train Videotape’s AI models. Instead, we:
Conduct regular internal research studies specifically to improve our models.
Use publicly available datasets with appropriate licensing.
IS OUR CUSTOMER SURVEY DATA USED BY THIRD-PARTY AI VENDORS TO TRAIN THEIR MODELS?
No.
Our contracts with AI vendors explicitly prohibit it. We only choose vendors that agree not to:
Use the data we send them for their model training.
Retain the data we send them for a significant period of time after it's been processed.
Share the data we send them with other parties.
Creative derivative or generative works from the data we send them.
WHAT TYPES OF DATA DO WE SEND TO THIRD-PARTY AI VENDORS?
We send only the data that’s minimally necessary to fulfill a specific task.
Data We May Send
Text transcripts, for sentiment analysis and theme extraction.
Anonymous text summaries, for insight generation.
Cropped or still video frames, used solely to confirm human presence or assist with fraud detection.
WHAT TYPES OF DATA DO WE NOT SEND TO THIRD-PARTY AI VENDORS?
We never transmit the following to third-party AI vendors:
Confidential customer information, including non-public information shared with Videotape..
Confidential customer media assets, including ads and other creative materials tested in survey.
Registered user account information, including names, emails, company affiliations.
Complete respondent video or audio recordings.
Respondent direct identifiers, including names and email addresses.
Financial data, including customer payment information and respondent compensation details.’
HOW IS CUSTOMER DATA SENT TO AND FROM EXTERNAL VENDORS?
In transit. All API calls use TLS 1.3 or higher for encryption in transit.
At rest. Data stored on third-party vendor servers is encrypted with AES-256.
Authentication. OAuth 2.0, API keys with scope limitations.
Videotape does not call external APIs directly from respondent devices.
All APIs are accessed through secure backend services, with keys stored and managed in a protected environment.
HOW DO WE STORE THE INPUTS AND OUTPUTS OF AI MODELS?
Storage Architecture
Infrastructure. Videotape's private cloud (Digital Ocean).
Segmentation. Isolated storage accounts with UUID-based object keys.
Access controls. Role-based permissions (RBAC) and authenticated user relationships.
Encryption. At-rest encryption for all AI data.
Retention. Customer data erased upon request.
Data Isolation
No intermingling of customer data.
One customer cannot see, query, or interact with another customer’s data.
External vendors never connect to our storage — any data shared is sent through Videotape’s secure backend with strict controls.
HOW ARE AI OUTPUTS USED?
Direct Research Applications
Report generation. Structuring and populating research reports.
Insight creation. Drafting summary insights for interpretation.
Pattern identification. Finding relationships between responses.
Visualization. Creating charts and multimedia assets to aid customer understanding.
Quality Assurance
All AI outputs intended for final reports are subject to a human review and validation process before delivery to customers.
Customers can request raw data alongside AI-processed insights.
Our reports are free of any AI-generated content that purports to represent, or simulate, a real human respondent.
HOW DO WE ENSURE RESPONDENTS ARE ASSESSED FAIRLY AND ACCURATELY BY AI?
Fraud detection and respondent bans are confirmed by human reviewers.
Respondent quality scores are audited.
Customers may request human review of any AI-generated insights.
WHAT IS OUR COMMITMENT TO SECURITY AND COMPLIANCE?
Our AI practices are designed to comply with global data protection and AI regulations, including GDPR, the California Privacy Rights Act (CPRA), and the principles of emerging AI laws in the US and EU. Our approach is grounded in our broader security program, which includes regular audits, access reviews, and adherence to our detailed Privacy Policy and Terms of Use
WHAT ARE YOUR RIGHTS?
Customers. You may request human review of AI-generated insights and export or delete their data subject to contractual limits.
Respondents: Your participation is voluntary. You can request Videotape to delete the metadata we have about you about you on our servers.
Please note that once your responses are included in a delivered customer report, we cannot retroactively remove them from that report, though we will delete all other data from our systems.
WHICH AI VENDORS DO WE USE?
OpenAI
Anthropic
Google Cloud Vertex AI
Questions or Concerns?
If you have any questions about this Agreement or would like to provide feedback, please contact us at legal@videotape.ai.