Free public instrument from GAGE
The AI Vendor Data Handling Ledger
As of September 2026 this ledger holds 31 records across 5 jurisdictions: 27 verified at a primary source, 0 reported by a secondary source, 0 announced with no document yet, 0 searched and absent, and 4 open questions. By kind: consumer tier 9, business tier 8, api and platform 10, open question 4.
- 31
- Records
- 27
- Verified at the source
- 0
- Announced or absent
- 4
- Open questions
5 jurisdictions, 103 primary sources
0 more reported, primary not reached
0 announced with no document, 0 searched and absent
Posed, sourced, not answered
As of 15 September 2026 the GAGE AI Vendor Data Ledger records 27 verified instruments, 0 reported at a secondary source, 0 announcements without a document, 0 absences and 4 open questions, across 5 jurisdictions. Counts are a floor, not a ceiling: an instrument the ledger has not found is not on it.
Why this ledger exists
Retention is a contract question, not a reputation question
The same vendor gives three different answers to what happens to your data, depending on which plan you are on and whether you reached it through an app or an API. A workplace policy that says which tools are allowed without saying which tier is allowed has not decided anything.
This ledger reads the vendor's own documents and records what each product tier says on ten dimensions, with the date the document was last changed. Where a document is silent, the record says silent, because a vendor that does not say is a fact a buyer needs as much as one that does.
Every row is fetched at its source where the source could be reached, and says so where it could not. The verdict language is the discipline: a reader learns five words once and never has to guess what a row claims.
Verified
The document exists. The ledger fetched it at its publisher and quotes it.
Reported, primary not reached
A reliable secondary source carries it, and the primary document could not be reached. Printed with this label, never as verified.
Announced, no document yet
A body has said it will act. No document exists yet, so the row records the statement and nothing more.
Absent
The ledger searched and found no instrument. The record says where it looked and when.
Open question
No settled answer exists. The ledger poses the question, links the live debate, and does not answer it.
The grid
What the documents state, and where they are silent
For each of the ten dimensions, how many product tiers on the ledger state a rule and how many are silent. A silent cell means the ledger read the vendor's documents and found no statement.
Used for training by default
26 stated, 1 silent.
Retention period
25 stated, 2 silent.
Deletion on request
23 stated, 4 silent.
Storage region and residency
21 stated, 6 silent.
Subprocessor list published
21 stated, 6 silent.
Human review of content
20 stated, 7 silent.
Opt out available
24 stated, 3 silent.
DPA available
16 stated, 11 silent.
Persistent memory across sessions
15 stated, 12 silent.
Last change to the terms
25 stated, 2 silent.
Side by side
Every jurisdiction, counted by verdict and by kind
United States
22 records
- Verified
- 22
- Reported
- 0
- Announced
- 0
- Absent
- 0
- Open question
- 0
7 consumer tier, 8 business tier, 7 api and platform.
European Union
1 record
- Verified
- 0
- Reported
- 0
- Announced
- 0
- Absent
- 0
- Open question
- 1
1 open question.
France
2 records
- Verified
- 2
- Reported
- 0
- Announced
- 0
- Absent
- 0
- Open question
- 0
1 consumer tier, 1 api and platform.
China
2 records
- Verified
- 2
- Reported
- 0
- Announced
- 0
- Absent
- 0
- Open question
- 0
1 consumer tier, 1 api and platform.
Global
4 records
- Verified
- 1
- Reported
- 0
- Announced
- 0
- Absent
- 0
- Open question
- 3
1 api and platform, 3 open question.
Figures of record
Every number on this ledger, with who measured it and when
45 figures, each one printed in the unit its publisher used, beside the publisher and the date it was true. Nothing here is summed across sources, converted between units, or forecast.
- 10 regions
Data residency regions offered. The OpenAI API and platform
OpenAI platform, data controls in the OpenAI platform, primary source, as of .
- 24 hours
Prompt cache expiry. The OpenAI API and platform
OpenAI platform, data controls in the OpenAI platform, primary source, as of .
- 30 days
Abuse monitoring log retention. The OpenAI API and platform
OpenAI platform, data controls in the OpenAI platform, primary source, as of .
- 30 days
Code execution container data retention. The Anthropic API and Claude Developer Platform
Anthropic platform docs, API and data retention, primary source, as of .
- 2 years
Retention of flagged data even under zero data retention. The Anthropic API and Claude Developer Platform
Anthropic platform docs, API and data retention, primary source, as of .
- 18 months
Default storage of conversation activity. Microsoft Copilot on a personal account
Microsoft, privacy FAQ for Microsoft Copilot, primary source, as of .
- 18 years
Minimum age for public content used in European training. Meta AI on Facebook, Instagram and WhatsApp
Meta newsroom, making AI work harder for Europeans, primary source, as of .
- 28 days
Automatic deletion of unused Copilot memory entries. GitHub Copilot Business and Enterprise
GitHub docs, Copilot memory, primary source, as of .
- 100 conversations
Conversations kept in Copilot Chat history. GitHub Copilot Business and Enterprise
GitHub docs, chat with Copilot on GitHub, primary source, as of .
- 28 days
Retention of Copilot Chat messages on GitHub. GitHub Copilot Business and Enterprise
GitHub docs, chat with Copilot on GitHub, primary source, as of .
- 30 days
Abuse detection retention for flagged traffic. Amazon Bedrock
AWS documentation, Amazon Bedrock abuse detection, primary source, as of .
- 30 days
Retention cap for models that require review. Amazon Bedrock
AWS documentation, Amazon Bedrock data retention, primary source, as of .
- 2 years
Retention of flagged data under zero data retention at one vendor. Does zero data retention survive an abuse investigation?
Anthropic platform docs, API and data retention, primary source, as of .
- 28 days
Automatic deletion of unused memory entries at one vendor. Who owns the memory record an assistant builds about you?
GitHub docs, Copilot memory, primary source, as of .
- 30 days
Advanced model prompt and response logging. The paid Gemini API and Vertex AI
Google Cloud, abuse monitoring, primary source, as of .
- 90 days
Abuse monitoring prompt log retention. The paid Gemini API and Vertex AI
Google Cloud, abuse monitoring, primary source, as of .
- 24 hours
Default in memory cache lifetime. The paid Gemini API and Vertex AI
Google Cloud, generative AI data governance, primary source, as of .
- 1 year
Retention of account data after deletion. The Mistral consumer assistant
Mistral AI privacy policy, primary source, as of .
- 5 years
Retention of civil identity data after account termination. The Mistral consumer assistant
Mistral AI privacy policy, primary source, as of .
- 30 days
Rolling abuse monitoring retention for standard API calls. The Mistral developer platform
Mistral AI privacy policy, primary source, as of .
- 30 days
Deletion after a person deletes conversations or the account. Grok on the consumer app and website
SpaceXAI privacy policy, published at x.ai, primary source, as of .
- 30 days
Deletion of Private Chat conversations. Grok on the consumer app and website
SpaceXAI privacy policy, published at x.ai, primary source, as of .
- 36 months
Ceiling of admin controlled retention for the Gemini app under Workspace. Gemini for Google Workspace
Google Workspace, generative AI privacy hub, primary source, as of .
- 90 days
Floor of the admin controlled retention range for Gemini in Workspace. Gemini for Google Workspace
Google Workspace, generative AI privacy hub, primary source, as of .
- 72 hours
Retention when the keep activity setting is off. The Gemini app on a personal Google account
Google, Gemini Apps privacy hub, primary source, as of .
- 3 years
Retention of human reviewed chats. The Gemini app on a personal Google account
Google, Gemini Apps privacy hub, primary source, as of .
- 18 months
Default auto deletion of Gemini Apps Activity. The Gemini app on a personal Google account
Google, Gemini Apps privacy hub, primary source, as of .
- 2,024 year
Date Copilot became a covered workload in the Product Terms residency commitments. Microsoft 365 Copilot, now named Microsoft Copilot
Microsoft Learn, data, privacy and security for Microsoft Copilot, primary source, as of .
- 7 years
Retention of trust and safety classification scores. Claude Free, Pro and Max
Anthropic privacy centre, how long do you store my data, primary source, as of .
- 2 years
Retention of inputs and outputs when a chat is flagged. Claude Free, Pro and Max
Anthropic privacy centre, how long do you store my data, primary source, as of .
- 30 days
Back end deletion after a person deletes a chat. Claude Free, Pro and Max
Anthropic privacy centre, how long do you store my data, primary source, as of .
- 5 years
Retention in model training pipelines when training is allowed. Claude Free, Pro and Max
Anthropic privacy centre, how long do you store my data, primary source, as of .
- 7 years
Retention of trust and safety classification scores. Claude for Work, Team and Enterprise
Anthropic privacy centre, how long do you store my organisation's data, primary source, as of .
- 2 years
Retention of flagged inputs and outputs. Claude for Work, Team and Enterprise
Anthropic privacy centre, how long do you store my organisation's data, primary source, as of .
- 30 days
Back end deletion of commercial inputs and outputs. Claude for Work, Team and Enterprise
Anthropic privacy centre, how long do you store my organisation's data, primary source, as of .
- 30 days
Default retention of API inputs and outputs. The Anthropic API and Claude Developer Platform
Anthropic privacy centre, how long do you store my organisation's data, primary source, as of .
- 30 days
Storage of prompts and output under grounding with Google Search. The Gemini API on the unpaid tier
Google, Gemini API additional terms of service, primary source, as of .
- 30 days
Deletion of personal data from OpenAI systems. ChatGPT Free, Plus and Pro
OpenAI privacy policy, primary source, as of .
- 30 days
Removal of deleted or unsaved conversations. ChatGPT Business, the plan formerly called Team
OpenAI enterprise privacy, primary source, as of .
- 30 days
Removal of deleted conversations. ChatGPT Enterprise and ChatGPT Edu
OpenAI enterprise privacy, primary source, as of .
- 30 days
Window to object to a new subprocessor. ChatGPT Business, the plan formerly called Team
OpenAI data processing addendum, primary source, as of .
- 30 days
Objection window after notice of a new subprocessor in one vendor addendum. Is a subprocessor change notice period actually enforceable?
OpenAI data processing addendum, primary source, as of .
- 30 days
Automatic deletion of logged prompts and generations. The Cohere platform and API
Cohere enterprise data commitments, primary source, as of .
- 30 days
Post termination deletion window on request. The Cohere platform and API
Cohere software as a service agreement, primary source, as of .
- 30 days
Automatic deletion of business inputs and outputs. The Grok API and enterprise platform
SpaceXAI enterprise FAQs, published at x.ai, primary source, as of .
The ledger
Every record, newest first
31 records. Each row opens a page carrying the answer, the verdict and what it means, the key facts, the figures with their sources, what it changes for a buyer or a workplace ai policy, and the sources it was verified against.
Answers
What people ask the AI Vendor Data Ledger
Which AI products train on what I type by default?
The training_use dimension on every record answers this for its tier as of September 2026. On the consumer tiers, 9 of 9 records carry a document that states training by default unless the person opts out; the record pages quote the sentence. Business and API tiers state the opposite default where the ledger could read it.
Which business tiers offer a data processing addendum?
8 of 8 business tier records state a DPA on the vendor's own page: GitHub Copilot Business and Enterprise; Cursor and its privacy mode; Perplexity enterprise and the Sonar API; Microsoft 365 Copilot, now named Microsoft Copilot; Gemini for Google Workspace; Claude for Work, Team and Enterprise; and more. A DPA is the document that makes retention and subprocessor promises enforceable rather than descriptive.
Where is my data stored?
The region dimension records what each document says about storage and residency. 6 records on the ledger are silent on it, which for a buyer in European Union is the question that decides whether the tier can be used at all.
How does the ledger handle a vendor changing its terms?
Every record carries the date the vendor's document was last changed as a dimension, and the record's own last verified date beside it. The ledger is re read every Monday and the same day a vendor announces a change; a change moves the record's date and appears on the changelog.
What happens to what a person types on a free or personal plan?
9 records of kind consumer tier are on the AI Vendor Data Ledger as of September 2026: Perplexity on the free and paid consumer plans; The DeepSeek consumer app; The Mistral consumer assistant; Grok on the consumer app and website; Meta AI on Facebook, Instagram and WhatsApp; Microsoft Copilot on a personal account; and 3 more. Each has its own page with the verdict, the facts, the sources it was checked against and the date.
What happens to what a team or an enterprise puts in?
8 records of kind business tier are on the AI Vendor Data Ledger as of September 2026: GitHub Copilot Business and Enterprise; Cursor and its privacy mode; Perplexity enterprise and the Sonar API; Microsoft 365 Copilot, now named Microsoft Copilot; Gemini for Google Workspace; Claude for Work, Team and Enterprise; and 2 more. Each has its own page with the verdict, the facts, the sources it was checked against and the date.
What happens to what a developer sends through the API?
10 records of kind api and platform are on the AI Vendor Data Ledger as of September 2026: The Cohere platform and API; Amazon Bedrock; The DeepSeek open platform and API; The Mistral developer platform; The Grok API and enterprise platform; Azure OpenAI Service and models sold by Azure in Microsoft Foundry; and 4 more. Each has its own page with the verdict, the facts, the sources it was checked against and the date.
What has nobody settled?
4 records of kind open question are on the AI Vendor Data Ledger as of September 2026: Who owns the memory record an assistant builds about you?; Does zero data retention survive an abuse investigation?; Is a subprocessor change notice period actually enforceable?; Does a promise not to train on your data cover the safety classifiers?. Each has its own page with the verdict, the facts, the sources it was checked against and the date.
What do the verdicts mean?
Verified: the document exists and the ledger fetched it at its publisher. Reported: a reliable secondary source carries it and the primary could not be reached. Announced: a body said it will act and no document exists. Absent: the ledger searched and found nothing, and the search is written into the record. Open: a question nobody has settled, posed and not answered.
How current is the AI Vendor Data Ledger?
Every record carries the date it was last verified; the ledger as a whole was last verified 15 September 2026 and holds 31 records with 103 primary sources. A change moves the record's own date and appears on the changelog, so a reader who cited a row can see whether it moved.
Every surface
Cut the ledger the way you need it
By jurisdiction
By kind
Every record page
- AVD-2026-0031: Who owns the memory record an assistant builds about you?
- AVD-2026-0030: Does zero data retention survive an abuse investigation?
- AVD-2026-0029: Is a subprocessor change notice period actually enforceable?
- AVD-2026-0028: Does a promise not to train on your data cover the safety classifiers?
- AVD-2026-0027: The Cohere platform and API
- AVD-2026-0026: Amazon Bedrock
- AVD-2026-0025: GitHub Copilot Business and Enterprise
- AVD-2026-0024: Cursor and its privacy mode
- AVD-2026-0023: Perplexity enterprise and the Sonar API
- AVD-2026-0022: Perplexity on the free and paid consumer plans
- AVD-2026-0021: The DeepSeek open platform and API
- AVD-2026-0020: The DeepSeek consumer app
- AVD-2026-0019: The Mistral developer platform
- AVD-2026-0018: The Mistral consumer assistant
- AVD-2026-0017: The Grok API and enterprise platform
- AVD-2026-0016: Grok on the consumer app and website
- AVD-2026-0015: Meta AI on Facebook, Instagram and WhatsApp
- AVD-2026-0014: Azure OpenAI Service and models sold by Azure in Microsoft Foundry
- AVD-2026-0013: Microsoft 365 Copilot, now named Microsoft Copilot
- AVD-2026-0012: Microsoft Copilot on a personal account
- AVD-2026-0011: The paid Gemini API and Vertex AI
- AVD-2026-0010: The Gemini API on the unpaid tier
- AVD-2026-0009: Gemini for Google Workspace
- AVD-2026-0008: The Gemini app on a personal Google account
- AVD-2026-0007: The Anthropic API and Claude Developer Platform
- AVD-2026-0006: Claude for Work, Team and Enterprise
- AVD-2026-0005: Claude Free, Pro and Max
- AVD-2026-0004: The OpenAI API and platform
- AVD-2026-0003: ChatGPT Enterprise and ChatGPT Edu
- AVD-2026-0002: ChatGPT Business, the plan formerly called Team
- AVD-2026-0001: ChatGPT Free, Plus and Pro
Take the data
The whole dataset, free, in two formats
Licensed CC BY 4.0. Use it in an article, a paper, a slide or a product. The only condition is attribution, and the citation page gives you the line to paste.
- ledger.jsonEvery field of every record, the shape documented on the data page.
- ledger.csvOne row per record, figures and facets flattened, for a spreadsheet or a stats package.
How the ledger is built, what the verdicts mean, and what the gate refuses: the method page. Every change, dated: the changelog. The kinds on the shelf: consumer tier, business tier, api and platform, open question. Something missing or wrong is a bug, and we want to hear about it. Before you paste anything into any of them, the Paste Test decides whether you should. Which of these tiers your people are already using is the Shadow AI Self-Audit. The policy that names an allowed tier comes from the AI Acceptable Use Policy Builder.
Cite this page
Free to reuse under CC BY 4.0, with attribution.
- In a sentence
- According to the GAGE AI Vendor Data Ledger (as of 15 September 2026), ai vendor data ledger.
- APA
- GAGE (Global Academy of Generative-AI Education). (2026). AI Vendor Data Ledger. AI Vendor Data Ledger. Retrieved 15 September 2026, from https://www.gage.academy/tools/ai-vendor-data-ledger
- MLA
- "AI Vendor Data Ledger." AI Vendor Data Ledger, GAGE (Global Academy of Generative-AI Education), 15 September 2026, https://www.gage.academy/tools/ai-vendor-data-ledger.
- Chicago
- GAGE (Global Academy of Generative-AI Education). "AI Vendor Data Ledger." AI Vendor Data Ledger. Last modified 15 September 2026. https://www.gage.academy/tools/ai-vendor-data-ledger.
- Permalink
- https://www.gage.academy/tools/ai-vendor-data-ledger
Last updated . Every record re verified . The ledger is checked weekly, every Monday, and the same day for any terms change a vendor announces.
10 regions
GAGE briefings tell you which AI regulation deadlines are coming, what they actually require of you, and when a program opens.