Quick look · a two-minute summary for reviewers

Kami and AI at a glance

The short answers to the questions every AI review asks, across Kami App, Kami Companion, Book Creator and Kami Coach.

Do students talk to an open AI chatbot?

Kami sends each request on the student's behalf, under Kami's rules.

Is student data used to train AI?

Never

No model training and no ad targeting.

Which model?

Gemini 3.5 Flash

Google Gemini, selected and managed by Kami.

Where is it processed?

US only

Google Vertex AI, US regions.

Can the district turn AI off?

Yes

In the Kami Admin Dashboard, for everyone or specific users.

Blog · 10-minute read · 15-minute video

Making sense of classroom AI in Kami

Plain-language definitions, where student work goes, and who holds the controls. Includes a one-pager and white paper to take to your board.

Read the blog

Each product in one line

Product	How it uses AI	Student chat?	Can be turned off
Kami App	Accessibility tools (explain, summarize, translate, relevel, read aloud) and teacher tools (Questions AI, suggested feedback)	No	Yes
Kami Companion	Explain, summarize and adjust in an assistive toolbar. Translate uses Google Translate, not generative AI	No	Yes
Book Creator	One AI-enabled feature: drafting image descriptions (alt text). Partner apps such as Adobe and Canva have their own AI	No	Yes
Kami Coach (beta)	AI coaching and hints inside teacher-assigned math activities	Inside the activity only	Yes

Independently reviewed and tested

Every new AI feature goes through a privacy, security and student well-being impact assessment, bias testing and QA testing before release, and is monitored after launch.

Independent auditor

SOC 2 Type II

Security controls audited over time

1EdTech

TrustEd Apps Data Privacy Certified

Student data privacy certification

Common Sense Media

Common Sense Privacy Seal

Independent privacy evaluation

Safer Technologies 4 Schools

ST4S

Security and privacy assessment for schools

CAST

UDL Guidelines 3.0

Kami App certified for Universal Design for Learning

Kami

EU AI Act declaration

Published declaration for Kami's AI features

Three ways to turn AI off

District and school admins

Kami Admin Dashboard

Turn AI off across the district, or for specific users.

Teachers

Assignment and class settings

Choose which tools each student can use, and when.

Kami

On request

Your Kami contact can switch AI off for your whole account.

Need more detail?

The AI task force view has every feature, the model behind it, how data flows and our security documents. The teacher view explains what students see in class.

For teachers

AI in your Kami classroom

Which Kami tools use AI, what your students see when they use them, and how you decide what's on.

Can my students chat with an AI?

They click a tool, such as Explain. There is no open chat box.

Is my students' work used to train AI?

Never

Their work stays inside Kami.

Who decides which tools are on?

You and your admin

Your district sets the limits. You choose within them.

Watch in 15 minutes

See these tools in action

In Making sense of classroom AI in Kami, Dr. Laurel Aguilar-Kirchhoff and David Hotler explain the definitions and demo the tools at classroom level.

Watch and read

Tools your students can use

Students select some text or part of an image, then pick a tool. Kami sends the request for them and shows the answer right in the document.

ExplainPuts selected text or an image into plain language.Kami App, Companion

SummarizePulls out the key ideas.Kami App, Companion

TranslateShows the text in another language.Kami App (AI), Companion (Google Translate)

Relevel and AdjustRewrites text so it's easier to read.Kami App, Companion

Read aloudReads text out loud, including text inside images.Kami App, Companion

Math coachingHints inside a math activity you assign.Kami Coach (beta)

Tools for you

Questions AICreates questions from your content, or turns a worksheet into interactive questions.Kami App

Suggested feedbackSuggests comments on student work. You review and edit before students see anything.Kami App

Image descriptionsDrafts alt text for images so screen reader users know what's there.Book Creator

Predictive text in Kami is rules-based. It doesn't use AI.

What you control

Which tools are on

Turn individual tools on or off for an assignment or class.

Which students

Give a tool to the students who need it, such as read aloud for a student with an IEP.

Partner apps

In Book Creator, switch off partner apps such as Adobe and Canva, which have their own AI.

If your district has turned an AI tool off in the Admin Dashboard, it won't appear for you or your students.

Why these tools matter

For many students these are accessibility supports. Read aloud lets a student who reads below grade level join the lesson. Translate makes the science unit reachable for a multilingual learner this week. Some of these supports are written into IEPs and 504 plans.

How Kami keeps AI answers safe and accurate

Starts from your contentTools work on the text or image your student selects, not on open-ended questions.

Written for learningEvery request carries Kami's education-focused instructions and safety rules.

Tested before releaseKami tests AI tools for bias, accuracy and age-appropriate answers before they reach your class.

You stay in chargeYou review suggested feedback and questions before students see them, and can report anything that looks wrong.

If a parent asks

"Kami's AI tools help students read, understand and translate their classwork. Students don't chat with an AI. They pick a tool and Kami shows the answer in their document. Their work is never used to train AI, and our district decides which tools are switched on."

For more detail, share the Kami AI quick look or the blog post Making sense of classroom AI in Kami with them.

For AI task forces, technology directors and privacy reviewers

Kami AI: the full picture

Every AI feature across Kami App, Kami Companion, Book Creator and Kami Coach, with the model, provider, data flow, safety testing and controls your review needs.

Start here · blog, video, one-pager and white paper

Making sense of classroom AI in Kami

The background to this page: how Kami defines generative, traditional, AI-powered and AI-enabled tools, and why those definitions matter when writing an AI policy. The white paper is built for vendor questionnaires and board presentations.

Read the blog

Definitions we use

These come from Making sense of classroom AI in Kami, which explains each one in more depth.

Generative AI

Creates new text from a prompt, using a large language model (LLM). Most classroom AI policies are written for this.

Traditional AI

Predicts, classifies or translates from things that already exist, such as Google Translate or neural reading voices.

AI-powered

AI is how the feature works. Turn the AI off and the feature stops.

AI-enabled

The feature works without AI, and AI enhances this feature. Today: Book Creator image descriptions.

Feature inventory

Rules-based, no AI model

Can be turned off

Not AI

Book Creator

Image descriptions (alt text)

AI-enabled

Drafts a description of an image for screen reader users. Descriptions can still be written by hand.

Who uses it

Teachers & students

Model / provider

Gemini 3.5 Flash (Google Vertex AI)

Can be turned off

Admin & teacher

Partner apps (Adobe, Canva)

AI-enabled

Partner apps include their own AI features, run by those partners.

Who uses it

Teachers & students

Model / provider

Partner’s own AI

Can be turned off

Admin & teacher

Kami Coach (beta)

AI coaching in math activities

Generative AI

Gives students hints and coaching inside a teacher-assigned math activity, and shows teachers how students are thinking.

Who uses it

Students & teachers

Model / provider

Gemini 3.5 Flash (Google Vertex AI)

Can be turned off

Admin & teacher

How student data flows

Kami acts as the intermediary. When a user picks an AI tool, Kami builds the request with education-focused instructions and safety rules, sends it to Google Vertex AI, and shows the answer inside the tool.

Not used for trainingStudent and teacher content is never used to train AI models.

No personal information to AIKami does not send students' personal information to AI models.

No advertisingNo ad targeting and no selling of data.

Separate from Google WorkspaceKami's AI runs on Kami's managed infrastructure. Your district's own Gemini settings are unaffected.

Short-lived at GoogleGoogle may hold request text for up to about two days for abuse monitoring, then it is removed.

Admin controls

AI features are on by default. Your district decides what stays on.

District and school admins

Kami Admin Dashboard

Turn AI features off across the whole district
Or turn them on and off for specific users

Teachers

Assignment and class settings

Choose which tools are available, for which students, and when
In Book Creator, switch off partner apps such as Adobe and Canva

Kami

Account-wide, on request

Kami can switch AI off across your district's account
Useful while your task force completes its review

Before you switch everything off

Read aloud, translate, explain and relevel are how many students reach grade-level material, and some are written into IEPs and 504 plans. Turning a feature off removes the support along with the AI. You can keep the supports students need and switch off only what your policy covers.

How Kami answers common review criteria

The questions AI vetting rubrics usually cover, with Kami's answer to each.

Student interaction

No open chatbot in Kami App, Companion or Book Creator. Students use fixed tools; Kami builds every request. Kami Coach coaching happens inside teacher-assigned math activities.

Human oversight

Teachers choose which tools each student can use. Suggested feedback and generated questions go to the teacher, who reviews and edits them before students see them.

Data privacy

No training on student data, no ad targeting. Covered by Kami's Privacy Policy and your data privacy agreement, in line with FERPA and COPPA.

Data residency

AI requests are processed on Google Vertex AI in US regions only.

Transparency

The model and provider are published here. AI-powered tools are labeled in the product so users can see which tools use AI.

Safety and accuracy

Requests carry Kami's education-focused instructions and safety rules. Kami's QA team tests outputs against a defined list of topics, test prompts and scripts before release, and monitors performance after launch.

Bias and fairness

Kami runs bias testing across user groups and learning contexts, following Google's Responsible AI Practices, as part of pre-release testing.

Hallucinations

Tools work on content the user selects, Kami's backend constrains inputs and structures outputs, and there is no open-ended chat. Teacher-facing outputs are reviewed by the teacher before students see them.

Curriculum and pedagogy

Instructions are education-focused and pedagogically grounded. Teachers keep control of instructional decisions, and tools are designed with educators.

District control

Admins turn AI off district-wide or for specific users in the Kami Admin Dashboard. Kami can also switch AI off for the whole account on request.

Accessibility and equity

Most AI features are accessibility supports. Kami App is CAST UDL Guidelines 3.0 certified.

Independent assurance

SOC 2 Type II from an independent auditor, 1EdTech TrustEd Apps Data Privacy Certified, Common Sense Privacy Seal, ST4S, and a published EU AI Act declaration. Kami's AI practices are subject to internal and third-party audits.

Testing, governance and independent review

How Kami checks AI features before and after release, as set out in Kami's Responsible AI and Ethics Commitment.

Governance

A cross-functional ethics review team (product, engineering, data privacy) reviews every new AI feature
Impact assessments before release cover privacy, security, and student learning and well-being
Kami can override or disable any AI feature that causes harm

Bias testing

Tested across different user groups and learning contexts
Follows Google's Responsible AI Practices for fairness
Repeated when models change

Accuracy and hallucinations

QA test scripts on a defined list of topics and prompts
Performance and reliability tested in simulated and real classroom settings
Ongoing monitoring for performance drops and unintended results

Designed to limit errors

Tools work on the content the user selects
Kami's backend constrains inputs and structures outputs
No open-ended chat, and AI features are labeled

Curriculum and pedagogy

Education-focused, pedagogically grounded instructions on every request
Teachers keep control of instructional decisions
Kami App is CAST UDL Guidelines 3.0 certified

Independent review

SOC 2 Type II from an independent auditor
1EdTech, Common Sense and ST4S privacy and security reviews
Internal and third-party audits of AI practices

Found an AI answer that looks wrong or inappropriate? Teachers and students can report it to Kami, and Kami investigates and fixes the cause.

Certifications and documents

Independent auditor

SOC 2 Type II

Report available under mutual NDA from security@kamiapp.com

1EdTech

TrustEd Apps Data Privacy Certified

Student data privacy certification

Common Sense Media

Common Sense Privacy Seal

Independent privacy evaluation

Safer Technologies 4 Schools

ST4S

Security and privacy assessment for schools

CAST

UDL Guidelines 3.0

Kami App certified

Regulation

FERPA, COPPA, UK and EU GDPR

Data processing agreements in the Security Hub

Security HubSOC 2, DPAs and EU AI Act declaration

Responsible AI and Ethics CommitmentGovernance, testing and human oversight

EU AI Act declarationKami's declaration for its AI features (PDF)

SubprocessorsWho processes data on Kami's behalf

Privacy PolicyIncluding Kami's use of AI

Making sense of classroom AI in KamiBlog, 15-minute video, one-pager and white paper

Kami's AI commitments

Consider humans first · Design AI for all users · Develop the most secure AI platform · Uphold industry standards and compliance · Foster transparency

Technical questions

Which models and providers does Kami use?

Generative features run on Google Gemini 3.5 Flash through Google Vertex AI. Companion's Translate uses Google Translate. Natural reading voices use Google Cloud Text-to-Speech, Kokoro (an open-source voice engine) or the device's own voices.

Is Kami's AI the same as our Google Workspace Gemini?

No. It is separate. Changing Gemini settings in Workspace does not affect Kami, and Kami's AI does not access your Workspace data.

Does Google keep the data?

Under Kami's commercial agreement, Google does not use the data to train models. Google may hold request text for up to about two days for abuse monitoring, then it is removed.

Are any features running AI in the background?

Yes, one. Kami App can read a document to suggest its grade, subject, publisher or course and a short summary, to help organize it. No student interacts with it.

Do you test for bias and hallucinations?

Yes. Kami runs bias testing across user groups and learning contexts, and tests accuracy and reliability with defined topics, prompts and scripts before release. Features are designed to limit errors: tools work on content the user selects, outputs are structured by Kami's backend, and teachers review teacher-facing outputs.

Has your AI been reviewed by a third party?

Kami's AI practices are subject to internal and third-party audits under our Responsible AI and Ethics Commitment. Kami's security controls are audited for SOC 2 Type II, and our privacy practices are certified by 1EdTech (TrustEd Apps), Common Sense and ST4S.

How do you change models?

Kami tests new model versions against its QA test set before moving to them, and updates this page when the model changes.

Can we see your SOC 2 report?

Yes. Your Kami contact can share the SOC 2 Type II report under NDA. Other documents are in the Security Hub.