Skip to content

Latest commit

 

History

History
1610 lines (1172 loc) · 28.2 KB

File metadata and controls

1610 lines (1172 loc) · 28.2 KB

Changelog

2026-07-27T19:29:17.794Z

CSV

SHA

dcbabb118a179d20a0e174967bee0833a876923ce1d9722a59dbd9d1ee8d2a45

canvasa11ycheckertablecaptions

model.name

- "Haiku 3"
+ "Haiku 3, Haiku 4.5"

revision

- "2026.02.19"
+ "2026.07.27"

canvasa11ycheckeralttextgenerator

model.name

- "Haiku 3"
+ "Haiku 3, Haiku 4.5"

revision

- "2026.02.25"
+ "2026.07.27"

canvascoursetranslation

model.name

- "Haiku 3"
+ "Haiku 4.5"

revision

- "2025.09.10"
+ "2026.07.27"

igniteagent

revision

- "2026.07.07"
+ "2026.07.27"

rubricgenerator

model.description

- "Anthropic Claude models are provided via Amazon Bedrock Foundation Models (FMs)."
+ "Anthropic Claude models are provided via Instructure's in-house AI Platform."

model.name

- "Haiku 3"
+ "Haiku 3, Haiku 4.5"

revision

- "2025.10.09"
+ "2026.07.27"

canvasgradingassistance

revision

- "2026.04.24"
+ "2026.07.27"

discussioninsights

model.description

- "Anthropic Claude models are provided via Amazon Bedrock Foundation Models (FMs)."
+ "Anthropic Claude models are provided via Instructure's in-house AI Platform."

model.name

- "Haiku 3"
+ "Haiku 3, Haiku 4.5"

revision

- "2025.10.02"
+ "2026.07.27"

canvasdiscussionsummaries

model.description

- "Anthropic Claude models are provided via Amazon Bedrock Foundation Models (FMs)."
+ "Anthropic Claude models are provided via Instructure's in-house AI Platform."

model.name

- "Haiku 3"
+ "Haiku 3, Haiku 4.5"

revision

- "2025.09.12"
+ "2026.07.27"

itemauthoringassistance

model.name

- "Haiku 3"
+ "Haiku 4.5"

revision

- "2025.11.18"
+ "2026.07.27"

assessmentauthoringassistance

model.name

- "Haiku 3"
+ "Haiku 4.5"

revision

- "2025.11.18"
+ "2026.07.27"

quickreassess

model.name

- "Haiku 3"
+ "Haiku 4.5"

revision

- "2025.11.19"
+ "2026.07.27"

aiexperiences

model.description

- "Anthropic Claude models are provided via Amazon Bedrock Foundation Models (FMs)."
+ "Anthropic Claude models are provided via Instructure's in-house AI Platform."

model.name

- "Haiku 3"
+ "Haiku 3, Haiku 4.5"

revision

- "2025.11.25"
+ "2026.07.27"

portfolios

model.name

- "Haiku 3"
+ "Haiku 3, Haiku 4.5"

studytools

supportpandabot

2026-07-07T20:36:03.843Z

CSV

SHA

fe1ecece8a55ab7f4ab289e7c8e4e277559807a92eea039618e61a9f8aec9770

igniteagent

compliance.regions

- "Virginia, Oregon, Frankfurt, Dublin, Sydney"
+ "Virginia, Oregon, Frankfurt, Dublin, Sydney, Montreal"

revision

- "2026.04.24"
+ "2026.07.07"

2026-07-07T20:36:03.843Z

CSV

SHA

fe1ecece8a55ab7f4ab289e7c8e4e277559807a92eea039618e61a9f8aec9770

igniteagent

compliance.regions

- "Virginia, Oregon, Frankfurt, Dublin, Sydney"
+ "Virginia, Oregon, Frankfurt, Dublin, Sydney, Montreal"

revision

- "2026.04.24"
+ "2026.07.07"

2026-06-23T17:40:09.483Z

CSV

SHA

17bda9fb3866cda71562613104200c295d62d17b80d4e06cc608c4b1171545fa

careerassistant

model.name

- "Claude 3.5 Haiku, Cohere multi-language v3"
+ "Claude 3 Haiku, Cohere multi-language v3"

2026-04-24T12:19:25.704Z

CSV

SHA

f4ad49ebae33ce95d33c1b2be8e0b5d0c4a7e0e9861a6454ec25767b39792685

igniteagent

compliance.piiDescription

- "If requested during the the chat, user, course, and student identifiers or metadata may be shared with the model."
+ "If requested during the chat, user, course, and student identifiers or metadata may be shared with the model."

revision

- "2026.04.22"
+ "2026.04.24"

canvasgradingassistance

compliance.regions

- "Global"
+ "Virginia, Oregon, Dublin, Frankfurt, Sydney"

compliance.regionsDescription

- "Available in all regions except for Singapre and Canada "
+ ""

revision

- "2026.04.22"
+ "2026.04.24"

2026-04-24T12:03:17.179Z

CSV

SHA

9ce383cc4c4feba02aa7162783d376022b5de5fbcd039b28341de10d57ce1a91

igniteagent

compliance.regions

- "Virginia, Oregon"
+ "Virginia, Oregon, Frankfurt, Dublin, Sydney"

revision

- "2026.01.14"
+ "2026.04.22"

canvasgradingassistance

compliance.regionsDescription

- ""
+ "Available in all regions except for Singapre and Canada "

model.name

- "Haiku 3"
+ "Haiku 4.5"

revision

- "2025.10.22"
+ "2026.04.22"

2026-04-14T17:32:10.363Z

CSV

SHA

46cc42abb4c2e74ec317905d153603d119118905b4911989ee57d9dc51d7b49c

igniteagent

model.name

- "Haiku 3, Haiku 3.5, Haiku 4.5, Sonnet 3.7"
+ "Haiku 4.5"

2026-03-31T17:27:34.751Z

CSV

SHA

2515120f879504865c1146bab59482a3010ee53ecd166775e0b2d7b83b56a1b4

igniteagent

model.name

- "Haiku 3, Sonnet 3.7"
+ "Haiku 3, Haiku 3.5, Haiku 4.5, Sonnet 3.7"

2026-02-25T21:52:22.088Z

CSV

SHA

8e7bcd63477a5917ac64ce5bf6cd2e3ae10fa087955aa49bceda3fa6071651a2

canvasa11ycheckertablecaptions

feature.description

- "A button to genereate table captions (based on the contents and immediate context of the table) for tables that don't have one."
+ "A button to generate table captions (based on the contents and immediate context of the table) for tables that don't have one."

revision

- "2026.02.19"
+ "2026.02.25"

canvasa11ycheckeralttextgenerator

outputs.risks

- "While the model can fairly accurately desicribe what's on a picture, it might not calways convey what the picture is meant to convey in the given context. "
+ "While the model can fairly accurately describe what's on a picture, it might not always convey what the picture is meant to convey in the given context. "

revision

- "2026.02.19"
+ "2026.02.25"

2026-02-25T19:27:11.739Z

CSV

SHA

7bde9b66ebce018ce8279def8dd1cf998ad0ecdd4c00cd4f3600700beb696f39

canvasgradingassistance

compliance.retention

- "Transactional data is retained for the life of the requestCanvas"
+ "Transactional data is retained for the life of the request."

canvasa11ycheckertablecaptions

feature.name

- "Table caption generation"
+ "Course Accessibility Checker: Table caption generation"

canvasa11ycheckeralttextgenerator

feature.name

- "Alt text generation"
+ "Course Accessibility Checker: Image alt text generation"

2026-01-15T16:20:16.119Z

CSV

SHA

efe50b60419bf433e14b99ce3063c634256ec55219c99a594fbd0dff851a31fa

canvasinboxtranslation

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

canvascoursetranslation

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

igniteagent

linkTitle

- undefined
+ "AI Privacy Policy"

linkURL

- undefined
+ "https://www.instructure.com/policies/artificial-intelligence-privacy"

model.description

- "Anthropic Claude models are provided via Amazon Bedrock Foundation Models (FMs)."
+ "Anthropic Claude models are provided via Instructure's in-house AI Platform."

outputs.guardrails

- "Access to the agent is limited to users with an Admin- or Teacher-based role. Data access and functions are scoped to the permissions available to the chat user."
+ "Access to the Agent is limited to users with an Admin- or Teacher-based role. Data access and functions are scoped to the permissions available to the chat user."

outputs.humanDescription

- "The agent only takes action based on human requests, and all write actions must be confirmed by the user."
+ "The Agent only takes action based on human requests, and all write actions must be confirmed by the user."

revision

- "2025.09.10"
+ "2026.01.14"

rubricgenerator

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

canvasgradingassistance

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

discussioninsights

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

smartsearch

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

askyourdata

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

canvasdiscussionsummaries

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

itemauthoringassistance

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

assessmentauthoringassistance

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

quickreassess

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

conversionalignment

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

bcealttext

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

aiexperiences

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

portfolios

linkTitle

- undefined
+ ""

linkURL

- undefined
+ ""

2025-12-23T17:09:13.054Z

CSV

SHA

520bd4c9fe948035a2332931643814abe1a4826218488c9236f1b703d7772108

canvasgradingassistance

compliance.retention

- "Transactional data is retained for the life of the request"
+ "Transactional data is retained for the life of the requestCanvas"

outputs.risks

- "Model may work better with certain kinds of assignment types or disciplines. The grader must remain alert to incorrect assumptions from the model."
+ "Grading Assistance does not flag unsafe content, verify facts or sources, enforce formatting or quantitative requirements, or reliably grade objective or single-answer questions. Results are based only on rubric criteria."

2025-11-25T17:08:58.760Z

CSV

SHA

f0f0627c5c8ea28f0528f8ee09ce51e66e0c9501e3d98ec403bda6257d4a58d4

aiexperiences

compliance

+ {
  "logging": "Logs data",
  "loggingDescription": "Prompt data, LLM responses, and student responses are logged.",
  "pii": "Not Exposed",
  "piiDescription": "",
  "regions": "Global",
  "regionsDescription": "",
  "retention": "All conversation data and the activity context created by the instructor is saved."
}

feature

+ {
  "description": "AI Experiences is a new activity type, allowing instructors to design and assess interactive chat experiences with AI.",
  "name": "AI Experiences"
}

group

+ "Canvas"

model

+ {
  "data": "Course, Student",
  "dataDescription": "The custom prompts that instructors provide to create the AI experience and student conversation with the AI Experience tool.",
  "description": "Anthropic Claude models are provided via Amazon Bedrock Foundation Models (FMs).",
  "name": "Haiku 3",
  "trained": "No"
}

outputs

+ {
  "guardrails": "Model responses are logged for quality assurance, and responses with low confidence are flagged \"Needs Review\" to encourage human intervention.",
  "human": "Yes",
  "humanDescription": "Instructors set the parameters for each conversation, and students are informed they are interacting with AI. The AI’s output is delivered as-is, without post-editing. Instructors have full access to the conversation transcript.",
  "outcomes": "New type of learning activity that incorporates conversational AI in the process of learning, allowing students to demonstrate understanding of a topic and engage in a new learning modality within the LMS.",
  "risks": "The LLM might not ask the right questions or assess the conversation correctly. ",
  "settings": "Yes"
}

permissions

+ "2"

revision

+ "2025.11.25"

uid

+ "aiexperiences"

2025-11-19T13:51:04.973Z

CSV

SHA

f62ad53ab4a0b498c59aa7a032648e344e0aa12cacd958c9a1f3594a0625d3bc

conversionalignment

compliance.retention

- "Standard execution, infrastructure, and other operational logs are retained for a reasonable period\nto enable monitoring and troubleshooting of underlying services."
+ "Standard execution, infrastructure, and other operational logs are retained for a reasonable period to enable monitoring and troubleshooting of underlying services."

outputs.outcomes

- "The intended outcomes are to streamline the process of aligning assessments with learning standards, save educators time, and\nensure consistency in assessment alignment."
+ "The intended outcomes are to streamline the process of aligning assessments with learning standards, save educators time, and ensure consistency in assessment alignment."

2025-11-19T13:49:14.970Z

CSV

SHA

02634d347b308c424937d4ed2e0621e548f9984ef3c860dedee9e79896af26c2

quickreassess

revision

- "2025.10.27"
+ "2025.11.19"

model.name

- "Claude Family Models"
+ "Haiku 3"

model.description

- ""
+ "Anthropic Claude models are provided via Instructure's in-house AI Platform."

conversionalignment

uid

+ "conversionalignment"

revision

+ "2025.11.19"

feature

+ {
  "name": "Standards Alignment for Quiz Conversion",
  "description": "When importing Canvas Quizzes to Mastery Connect, AI can be used to suggest aligned standards."
}

model

+ {
  "name": "Claude Family Models",
  "description": "Anthropic Claude models are provided via Amazon Bedrock Foundation Models (FMs).",
  "trained": "No",
  "data": "Course",
  "dataDescription": "The Canvas Quiz content and Mastery Tracker metadata (Subject, Objectives) is shared with the\nmodel."
}

compliance

+ {
  "retention": "Standard execution, infrastructure, and other operational logs are retained for a reasonable period\nto enable monitoring and troubleshooting of underlying services.",
  "logging": "Does not log data",
  "loggingDescription": "No user data is logged at any time.",
  "regions": "Virginia, Oregon",
  "regionsDescription": "",
  "pii": "Not Exposed",
  "piiDescription": ""
}

outputs

+ {
  "settings": "Yes",
  "human": "Yes",
  "humanDescription": "The aligned standards can be reviewed and adjusted by educators prior to publishing the content.",
  "guardrails": "AI suggestions are limited to the standards present in the linked tracker’s standard set. This ensures that only the intended audience (educators) can use the tool and only relevant standards are suggested.",
  "risks": "Expected risks include possible misalignment if AI suggestions are not carefully reviewed, reliance on the accuracy of the tracker’s standard set, and the need for educator oversight to ensure quality.",
  "outcomes": "The intended outcomes are to streamline the process of aligning assessments with learning standards, save educators time, and\nensure consistency in assessment alignment."
}

group

+ "Mastery"

permissions

+ "2"

2025-11-18T20:25:05.839Z

CSV

SHA

858e96391530102331964186a144034150fc6d61c0e84d916bd083975d3f99f4

itemauthoringassistance

uid

+ "itemauthoringassistance"

revision

+ "2025.11.18"

feature

+ {
  "name": "Question Authoring",
  "description": "Question Authoring Assistance for Quizzes streamlines quiz creation by generating questions based on your course materials. \n"
}

model

+ {
  "name": "Haiku 3",
  "description": "Anthropic Claude models are provided via Instructure's in-house AI Platform.",
  "trained": "No",
  "data": "Course",
  "dataDescription": "Source material provided by the user (e.g., Custom standards, course content, text input, file input) for question/item generation."
}

compliance

+ {
  "retention": "Data is not stored or reused by the model.",
  "logging": "Logs data",
  "loggingDescription": "Usage data is logged to improve the product.",
  "regions": "Global",
  "regionsDescription": "",
  "pii": "Not Exposed",
  "piiDescription": "PII in source material may be sent to the model if included by the user, but no PII is intentionally sent to the model."
}

outputs

+ {
  "settings": "Yes",
  "human": "Yes",
  "humanDescription": "Instructors can review, edit, regenerate, or provide additional instructions to refine generated items.",
  "guardrails": "",
  "risks": "Generated items may not perfectly capture all desired information, require refinement, or occasionally contain inaccuracies.\n",
  "outcomes": "Improved instructor efficiency and productivity"
}

group

+ "Canvas"

permissions

+ "2"

assessmentauthoringassistance

uid

+ "assessmentauthoringassistance"

revision

+ "2025.11.18"

feature

+ {
  "name": "Item Authoring",
  "description": "Item Authoring for Assessments streamlines assessment creation by generating questions based on your course materials. \n"
}

model

+ {
  "name": "Haiku 3",
  "description": "Anthropic Claude models are provided via Instructure's in-house AI Platform.",
  "trained": "No",
  "data": "Course",
  "dataDescription": "Source material provided by the user (e.g., Custom standards, course content, text input, file input) for question/item generation."
}

compliance

+ {
  "retention": "Data is not stored or reused by the model.",
  "logging": "Logs data",
  "loggingDescription": "Usage data is logged to improve the product.",
  "regions": "Global",
  "regionsDescription": "",
  "pii": "Not Exposed",
  "piiDescription": "PII in source material may be sent to the model if included by the user, but no PII is intentionally sent to the model."
}

outputs

+ {
  "settings": "Yes",
  "human": "Yes",
  "humanDescription": "Instructors can review, edit, regenerate, or provide additional instructions to refine generated items.",
  "guardrails": "",
  "risks": "Generated items may not perfectly capture all desired information, require refinement, or occasionally contain inaccuracies.\n",
  "outcomes": "Improved instructor efficiency and productivity"
}

group

+ "Mastery"

permissions

+ "2"

2025-11-18T19:52:02.235Z

CSV

SHA

6e5d47862429c5073e26beebe10100fbaa28615f7d76f7eaa4f7c94a36b7aaeb

canvasinboxtranslation

feature.name

- "Inbox Translation"
+ "Translations for Inbox"

canvascoursetranslation

feature.name

- "Discussions Translation"
+ "Translations for Discussions & Announcements"

rubricgenerator

feature.name

- "Rubric Creator"
+ "Generator for Rubrics"

discussioninsights

feature.name

- "Discussion Insights"
+ "Insights for Discussions"

smartsearch

feature.name

- "Smart Search"
+ "Search"

canvasdiscussionsummaries

feature.name

- "Discussion Summaries"
+ "Summaries for Discussions"

2025-11-04T17:07:14.773Z

CSV

SHA

7ca7919d381e049839aef8eebac865f13e60b71d8191fcd47d41db2e0be45d59

igniteagent

model.name

- "Haiku 3.5, Sonnet 3.5"
+ "Haiku 3, Sonnet 3.7"

2025-10-28T20:56:32.209Z

CSV

SHA

cb94462a64ddfdf50e03c51f9d7956ebfb09699f9cf426027e06888920f5d9c5

canvasgradingassistance

outputs.humanDescription

- "Grade suggestion is displayed and selected by default, but the instructor much make the final decision to edit or accept the suggestions."
+ "Grade suggestion is displayed and selected by default, but the instructor must make the final decision to edit or accept the suggestions."

quickreassess

uid

+ "quickreassess"

revision

+ "2025.10.27"

feature

+ {
+   "name": "Suggestions for Quick Reassess",
+   "description": "Allows the use of AI to generate a question derived from the original. These questions can be re-generated, edited, and rejected or approved before they are published to students."
+ }

model

+ {
+   "name": "Claude Family Models",
+   "description": "",
+   "trained": "No",
+   "data": "Other",
+   "dataDescription": "The original question stem used to generate derivatives"
+ }

compliance

+ {
+   "retention": "Standard execution, infrastructure, and other operational logs are retained for a reasonable period to enable monitoring and troubleshooting of underlying services.",
+   "logging": "Does not log data",
+   "loggingDescription": "",
+   "regions": "Virginia",
+   "regionsDescription": "",
+   "pii": "Not Exposed",
+   "piiDescription": ""
+ }

outputs

+ {
+   "settings": "Yes",
+   "human": "Yes",
+   "humanDescription": "Educators have access to tools that allow them to preview, regenerate, edit, or replace AI generated questions before they are published in an assessment.",
+   "guardrails": "Only educators can generate items and all AI generated items are in draft status until reviewed and approved.",
+   "risks": "AI generated questions may require additional review and editing to ensure accuracy, relevance, and alignment with intended learning objectives.",
+   "outcomes": "Educators can generate new standards-aligned questions based on existing items in order to quickly create diverse, tailored assessments that expand beyond the current Mastery Item Bank content, enhancing instructional flexibility and efficiency."
+ }

group

+ "Mastery"

permissions

+ "2"

2025-10-27T17:53:08.481Z

CSV

SHA

fbcefd273d8991abc9dbeef49b025502650f7decce5fcaced0657ef46983fd7a

portfolios

revision

- "2025.10.21"
+ "2025.10.27"

outputs.guardrails

- "Teacher access to the model is limited, it's only through the instructions of the portfolio/evidence. There is suspicious prompt detection built in for extra security. "
+ "Teacher access to the model is limited, it's only through the instructions of the portfolio/evidence. There is suspicious prompt detection built in for extra security."

2025-10-27T16:47:22.334Z

CSV

SHA

764a870a31eaef555d1956b203241631476ac610403d1db3ecca305c7a947000

canvasGradingAssistance

revision

- "2025.10.14"
+ "2025.10.22"

model.name

- "Haiku 3.5"
+ "Haiku 3"

outputs.guardrails

- "The Claude Sonnet model has inherent guardrails built in."
+ "The Claude Haiku model has inherent guardrails built in."

2025-10-21T20:13:52.750Z

CSV

SHA

fb0b079f1550d1302447434d4ccf7d126afa237f42e44946334365b4784e5741

portfolios

outputs.guardrails:

- "Teacher access to the model is limited, it's only through the instructions of the portfolio/evidence. There is suspicious prompt detection build in for extra security. "
+ "Teacher access to the model is limited, it's only through the instructions of the portfolio/evidence. There is suspicious prompt detection built in for extra security. "

2025-10-21T18:03:42.385Z

CSV

SHA

e4fde48757a6d375a20a822aecb74e7ac9f7d2bb43204243b5f3a6240882fb45

portfolios

revision:

- "2025.10.17"
+ "2025.10.21"

feature.description:

Old:

- "AI Assitant for teachers to improve portfolio level and evidence level instructions improvement, skill extraction and reflection question generation."
+ "AI Assistant for teachers to improve portfolio level and evidence level instructions improvement, skill extraction and reflection question generation."

outputs.guardrails:

- "Teachers acces to the model is limited, it's only through the intructions of the portfolio/ evidence. There is suspicious prompt detection build in for extra security."
+ "Teacher access to the model is limited, it's only through the instructions of the portfolio/evidence. There is suspicious prompt detection build in for extra security."

2025-10-21T17:35:42.207Z

CSV

SHA

4c97aaba0cddade6535f715cd3800d838ce62e2f6971babf3dc84f5329a00bc5

canvasinboxtranslation

model.description:

- "Anthropic Claude models are provided via Instructure's in-house AI Platform."
+ "AWS Translate model is provided via Instructure's in-house AI Platform."

2025-10-21T17:08:20.332Z

CSV

SHA

86553f2cc37930e62ff8c3af69c3c384a4d93320d855018058144f27d05d5fb5

canvascoursetranslation

uid

+ "canvascoursetranslation"

revision

+ "2025.09.10"

feature

+ {
+   "name": "Discussions Translation",
+   "description": "Translation of Discussion threads (\"Course AI Translation\" feature flag) across 10 languages."
+ }

model

+ {
+   "name": "Haiku 3",
+   "description": "Anthropic Claude models are provided via Instructure's in-house AI Platform.",
+   "trained": "No",
+   "data": "Course",
+   "dataDescription": "Announcement and Discussion prompts and replies"
+ }

compliance

+ {
+   "retention": "Data is not stored or reused by the model.",
+   "logging": "Does not log data",
+   "loggingDescription": "",
+   "regions": "Global",
+   "regionsDescription": "",
+   "pii": "Not Exposed",
+   "piiDescription": "PII in discussion replies may be sent to the model, but no PII is intentionally sent to the model."
+ }

outputs

+ {
+   "settings": "Yes",
+   "human": "Yes",
+   "humanDescription": "Untranslated content is available to review translations against",
+   "guardrails": "",
+   "risks": "Machine translation may not fully capture the meaning of the original message.",
+   "outcomes": "Improve participation for students who do not natively speak the language of instruction or other replies."
+ + }

group

+ "Canvas"

permissions

+ "2"

portfolios

uid

+ "portfolios"

revision

+ "2025.10.17"

feature

{
+   "name": "Portfolios",
+   "description": "AI Assitant for teachers to improve portfolio level and evidence level instructions improvement, skill extraction and reflection question generation."
+ }

model

+ {
+   "name": "Haiku 3",
+   "description": "Anthropic Claude models are provided via Instructure's in-house AI Platform.",
+   "trained": "No",
+   "data": "Other",
+   "dataDescription": "Custom instructions that teachers provide in their Evaluation portfolios for students."
+ }

compliance

+ {
+   "retention": "Data is not stored or re-used by the model",
+   "logging": "Logs data",
+   "loggingDescription": "Model evaluations and reply labels are logged for debugging and troubleshooting purposes.",
+   "regions": "Global",
+   "regionsDescription": "",
+   "pii": "Not Exposed",
+   "piiDescription": ""
+ }

outputs

+ {
+   "settings": "Yes",
+   "human": "Yes",
+   "humanDescription": "Teachers provide the input of the model (content) and the response by the model is not auto accepted. Teachers have the chance to modify the result every occasion.",
+   "guardrails": "Teachers acces to the model is limited, it's only through the intructions of the portfolio/ evidence. There is suspicious prompt detection build in for extra security. ",
+   "risks": "LLM might suggest not the right questions or skills, or not an improved instruction to the user's liking.",
+   "outcomes": "A user-friendly tool for a safe writing assistant, and features to speed up administrative work, and provide more reliable portfolios for students to work with."
+ }

group

+ "Canvas"

permissions

+ "2"