Tosa uses artificial intelligence selectively and transparently to help institutions deliver better assessments and support candidates more effectively — always under your control, and never at the expense of your data.
Our Approach to AI
AI on the Tosa platform is a tool for educators and administrators, not a replacement for human judgment. Every AI-powered feature is optional, configurable by your institution, and designed to protect the privacy of candidates and the integrity of assessments.
The principles below guide how we build, deploy, and limit AI across all Tosa services.
- 🔍 Transparency — We tell you when and how AI is being used within the platform.
- 🎛️ Your choice — All AI features are optional and can be turned off entirely at any time.
- 🔒 Privacy by design — Personal information — names, emails, identifiers — is never sent to AI systems.
- 🚫 No training on your data — Tosa never uses client or candidate data to train AI models or build datasets.
- ⚖️ Human oversight — AI assists — it does not decide. Final decisions remain with your institution.
- 📦 Data minimization — AI systems receive only the minimum information required to complete a task.
What AI Does on Tosa
When enabled by your institution, AI may support the following functions:
- 📊 Assessment reporting and competency analysis — AI can generate enhanced performance summaries and skill gap insights across candidate cohorts, using anonymized assessment data only.
- ✏️ Automated grading of open-ended questions — AI can score free-response answers based on rubrics you define. Institutions may require human review before grades are finalized.
- 💡 Question creation support — AI can suggest question titles, formative prompts, and structural improvements to help instructors build assessments more efficiently.
- 🔎 Question quality review — AI can flag clarity issues, spelling and grammar problems, and logical inconsistencies in answer options — supporting content quality assurance.
- 🖥️ Remote proctoring image analysis — Where remote proctoring is used, AI may assist with image analysis during exam supervision. Only non-identifiable images are processed, and your institution selects the AI provider used.
Choosing Your AI Provider
Your institution selects which AI provider powers the features you enable. Tosa currently supports the following approved generative AI providers: OpenAI (ChatGPT), Mistral AI, and Anthropic (Claude).
Vendor selection is your institution's decision. You can change your preferred AI provider through your platform configuration settings, or contact your Tosa administrator for assistance.
Data Protection and AI
Tosa's AI integrations are built around strict data protection principles consistent with our Privacy Policy.
- No personal data is shared with AI systems. Names, email addresses, passwords, and candidate identifiers are never transmitted to any AI provider.
- AI systems do not retain your data. Submitted content is processed in the moment and is not stored, reused, or repurposed by the AI provider.
- Your data is never used to train AI models. Isograd does not use client or candidate data — including assessment content, responses, or personal information — to train language models, develop algorithms, or build datasets.
- AI modules are isolated. AI services within Tosa are separated from marketing systems and analytics pipelines, and are only connected to your selected AI provider.
Controls Available to Your Institution
Client institutions have full control over AI on the platform. You can:
- Enable or disable AI features platform-wide or for individual services
- Choose which AI provider is used
- Request that all AI functionality be turned off completely at any time
To adjust your AI settings or ask questions about how AI is used in your account, please contact your Tosa representative or reach us via our Contact page.
This page summarizes Isograd's AI Governance Policy for the Tosa platform. For the complete policy document, please contact us. For information on how Tosa handles personal data more broadly, see our Privacy Policy.