Responsible AI Policy
Last updated: June 16, 2026
Our Commitment
Tightrope Media Systems is committed to building and using artificial intelligence (AI) responsibly in our products. We believe AI should expand what people can do, make content more accessible, and be used in the open: clearly, honestly, and ethically.
Scope
This policy applies to all Tightrope products that use AI or machine learning, including MediaScribe Live and Narrate features, as well as any future AI-enabled offerings. It covers both the AI that analyzes content (such as turning speech into text) and the generative AI that creates new content (such as writing audio descriptions and turning them into spoken narration).
The Two Kinds of AI We Use
To keep things clear, it helps to know that our products use AI in two different ways:
AI that analyzes. This AI listens to or looks at your content and turns it into something useful, like converting speech to captions, translating those captions into other languages, identifying who is speaking, or finding pauses in the audio. It works with what is already there.
AI that generates. This is "generative AI." It creates new content that did not exist before. In MediaScribe Narrate, generative AI looks at the video and listens to the audio, writes a short description of what is happening on screen, and produces a spoken voice track from that description.
Generative AI is powerful, but because it writes something new, it can occasionally get details wrong. That is why the safeguards in this policy, especially human review, matter.
Core Principles
1. Transparency
We are open about how our products use AI:
Clear communication. We tell you when our products use AI or machine learning, what those features do, and where they have limits.
No hidden AI. You are told about AI-powered features before and during use.
Labeled AI-generated content. Content created by generative AI, such as audio descriptions, is identified as AI-generated so no one mistakes it for human-authored narration.
Documentation. We explain how our AI features work in plain language that our customers can understand.
Open about service status. We publish real-time service status and incident history at https://www.mediascribe.ai/status.
2. Privacy and Data Protection
We protect customer data and respect privacy:
No training on customer data. Your audio, video, images, transcripts, and the descriptions we generate are NOT used to train AI or machine learning models.
Zero data retention with our AI providers. The third-party AI services we use do not retain, log, or train on the content we send them for processing.
Data residency. Customer content is processed and stored in data centers located in the United States.
Data minimization. We only collect and process the data needed for the service to work.
Data retention. Your content is stored only for the term of your subscription. After it is deleted, removal is completed within 90 days, including from backups.
Encryption. Data in transit is encrypted using industry-standard protocols (SSL/HTTPS).
3. Accuracy and Fairness
We work to make AI that performs well for everyone:
Diverse performance. We choose AI providers who show strong accuracy across different speakers, accents, and dialects.
Continuous improvement. We regularly update our products to take advantage of accuracy improvements.
Honest about limits. We clearly explain conditions where AI performance can drop, such as with poor audio, heavy accents, technical jargon, or fast-moving or visually complex video.
Watching for mistakes in generated content. Generative AI can occasionally describe something that is not there, miss an important visual detail, or word something awkwardly. We design Narrate so a person can catch and fix these before the content is shared.
Bias awareness. We recognize that AI systems can carry bias, and we work with partners who actively measure and reduce it.
4. Human Oversight
Our AI supports people; it does not replace human judgment:
Assistive by design. Our AI provides accessibility tools (captioning, translation, and describing) rather than making automated decisions about individuals.
A person reviews generated content before it is published. Audio descriptions created by Narrate are meant to be reviewed and approved by a person before distribution. Our editing tools let you play back, correct the wording, adjust timing, or flag a description before the accessible video goes out.
You stay in control. You can change, correct, and improve any AI- or machine-generated content.
5. Security
We maintain strong security practices:
Vetted partners. We work with AI service providers who hold recognized security certifications (such as ISO 27001 and SOC 2 Type II).
Regular testing. Our partners conduct regular penetration testing and security audits.
Incident response. We maintain procedures for responding to AI-related issues.
6. Appropriate Use
Our AI is built for beneficial purposes:
Accessibility focus. MediaScribe exists to improve accessibility for people who are deaf, hard of hearing, blind, low-vision, or who speak different languages.
Not high-risk. Our speech recognition and audio-description features are not designed to make decisions about people and do not fall into the high-risk or prohibited categories defined by the EU AI Act.
Responsible generative AI. Where we use generative AI, such as creating audio descriptions, we apply careful guardrails, label the output as AI-generated, and keep a person in the loop to review and approve the content before it is distributed.
Spotlight: MediaScribe Narrate
MediaScribe Narrate is our audio-description feature. It makes video accessible to blind and low-vision viewers by describing what is happening on screen. Here is how it uses AI, and how we keep it responsible:
It analyzes the video. Speech recognition creates an accurate transcript and finds the natural pauses in dialogue. Computer vision identifies meaningful things happening on screen, like a speaker gesturing toward a slide.
It generates a description. Generative AI writes a short, plain description of the important visual moments and fits each one into a pause so it never talks over the dialogue.
It creates the narration. Neural text-to-speech turns the written description into natural-sounding spoken audio.
A person reviews it. Before the accessible video is shared, a reviewer can listen to each description, edit the wording, adjust the timing, or flag anything that needs attention.
Because Narrate describes visual content rather than just transcribing words, we hold it to the extra safeguards in this policy: clear labeling, human review before publishing, and honesty about what it can and cannot do.
Our AI Partners
MediaScribe relies on a small set of established technology providers to power its AI features. We choose partners that meet strong security and privacy standards, process data in U.S. data centers, and operate under zero data retention, meaning they do not keep, log, or train on the content we send them. Each provider's role and security posture is summarized below.
Google (Gemini)
Role: Visual analysis of the video, identifying what is happening on screen.
Security posture: Google Cloud holds SOC 1, 2, and 3 reports, ISO 27001, and ISO 42001 (the international standard for responsible AI management), and meets U.S. FedRAMP High. Content processed through its enterprise AI services is not used to train Google's models.
Anthropic (Claude)
Role: Analyzes the transcript and writes the audio descriptions.
Security posture: Certified to SOC 2 Type II, ISO/IEC 27001:2022, and ISO/IEC 42001:2023 (the international standard for responsible AI management). API data is never used to train its models, and zero-data-retention handling means inputs and outputs are not stored after processing.
Amazon Web Services (AWS)
Role: Translation, text-to-speech narration (Polly), storage (S3), and video processing (MediaConvert) behind MediaScribe Live and Narrate.
Security posture: AWS maintains one of the broadest compliance programs in the industry, including SOC 1, 2, and 3, ISO 27001, PCI DSS, and FedRAMP authorization for U.S. government workloads.
Speechmatics
Role: The speech recognition (ASR) that turns spoken audio into accurate transcripts.
Security posture: A UK-based company subject to UK and EU GDPR, certified to ISO/IEC 27001:2022 and SOC 2 Type II, compliant with HIPAA, with industry-leading accuracy across diverse speakers. It does not use customer data to train its models.
Heroku (Salesforce)
Role: Secure application hosting for the MediaScribe platform.
Security posture: Operated by Salesforce under an ISO 27001-certified information security program (aligned to ISO 27017 and 27018 for cloud and personal-data protection), with SOC 1, 2, and 3 reporting.
Honest Communication About What Our AI Does
We believe in being straightforward about what our AI can and cannot do.
What our AI does well
Transcribes clear speech in supported languages, in real time
Translates into multiple languages
Identifies different speakers and adds punctuation and capitalization
Drafts plain-language descriptions of important on-screen visuals and places them in pauses in the dialogue
Turns those descriptions into natural-sounding spoken narration
Known limitations
Accuracy can drop with background noise, echo, or poor audio quality
Unfamiliar accents, dialects, or languages may reduce accuracy
Technical jargon, proper nouns, and acronyms may not be recognized correctly
Multiple people speaking at once can cause errors
Generated descriptions can occasionally misread a visual scene, miss a detail, or word something awkwardly, especially in fast-moving or visually complex video, which is why a person reviews them before publishing
What our AI does NOT do
Make decisions or recommendations about individuals. Our AI is an accessibility tool, not a decision-maker.
Learn from or retain your content. Your data is not used to train models and is deleted at the end of your subscription.
Replace human judgment. Generated descriptions are meant to be reviewed and approved by a person before they are shared.
Invent facts about people or events. Narrate describes what it observes on screen; it does not add commentary, opinions, or information that is not in the video.
Governance and Accountability
Policy ownership. This policy is owned by Tightrope Media Systems leadership and reviewed whenever we add a significant new AI feature.
Partner assessment. We evaluate our AI partners' practices, certifications, and commitments to responsible AI.
Customer feedback. We welcome feedback about AI performance and concerns, and we use it to improve.
Continuous learning. We stay informed about evolving AI regulations, best practices, and industry standards.
Education. We offer easy-to-understand resources that explain our AI features, their purpose, and the controls available to you.
Contact
For questions about this policy or our AI practices, please contact Tightrope Media Systems at sales@mediascribe.ai or 612-979-9719.
Last updated: June 2026