“This is certainly logically possible, though not well-explained elsewhere in Microsoft’s documentation, such as this which says (vaguely) that ”connected experiences [may] use machine learning services to … Analyze Data…” among others things.”
Our reading, the sentence that was checked:
Microsoft's documentation does not clearly explain whether Connected Services data is used to train its large language models, despite saying connected experiences may use machine learning to analyze data.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Update re: Microsoft and training data, paragraph 2.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-update-re-microsoft-and-training-0d292f-a2
We recorded 3 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2024-11-26. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
From the speaker
- not well-explained elsewhere in Microsoft’s documentation
- connected experiences [may] use machine learning services
- Analyze Data
Our interpretation
- The plain reading concerns whether Microsoft's documentation clearly explained how Connected Experiences content analysis relates to LLM training. It does not assert that Microsoft actually trained LLMs on this content.
- Assess the documentation publicly available when the post appeared on November 26, 2024. Documentation is continuously published, so the relevant observation period is that day's public documentation, rather than a reporting quarter.
- A clear explanation would explicitly address training use, or provide an unambiguous exclusion covering the relevant Connected Experiences data. A description of content analysis alone does not answer that question.
- The context includes consumer and commercial Microsoft 365 applications. A statement limited to organizational Copilot data does not establish the position for every Connected Experience.
Counts as supported if: Historical versions of the linked documentation and relevant accompanying privacy documentation show that they described content analysis without clearly explaining whether the relevant consumer and commercial Connected Experiences data could be used for LLM training.
Counts as contradicted if: Documentation demonstrably available before the post clearly states whether the relevant Connected Experiences data is used for LLM training, with coverage matching the consumer and commercial scope of the claim.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The November 28, 2024 capture describes machine-learning content analysis without expressly addressing LLM training, but it does not establish the page's wording before the November 26 post [e1]. Copilot documentation dated November 15 explicitly excludes specified organizational data from foundation-model training; its coverage does not establish the position for all consumer and commercial Connected Experiences [e2]. November 26 reporting records Microsoft's denial of training on Microsoft 365 customer data, but that statement does not establish what its documentation said [e3]. These items leave the documentation claim unresolved for its date and scope [e1–e3].
Limitations
- No verified pre-post capture or revision history establishes the linked page's wording on November 26, 2024.
- The packet does not establish what all applicable consumer and commercial privacy documentation said.
- Documentation clarity does not establish Microsoft's actual training practices.
Evidence (3 items)
Context only: Microsoft Learn, archived by the Internet Archive,
Some connected experiences use machine learning services to perform their function to help you accomplish a task.
Why it matters here: The archived page lists Analyze Data among machine-learning features and does not expressly address LLM training. This is consistent with Marcus's concern, but the verified capture is two days after the target date.
- Publisher
- Microsoft Learn, archived by the Internet Archive
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: Microsoft Learn, archived by the Internet Archive,
Prompts, responses, and data accessed through Microsoft Graph aren't used to train foundation LLMs, including those used by Microsoft 365 Copilot.
Why it matters here: The retrieved version bears this article date and clearly excludes the specified organizational Copilot data from training. It qualifies any sweeping claim that Microsoft supplied no explanation, but does not cover all consumer and commercial Connected Experiences data.
- Publisher
- Microsoft Learn, archived by the Internet Archive
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: How-To Geek,
Additionally, the Connected Services setting has no connection to how Microsoft trains large language models.
Why it matters here: Contemporary reporting records a clear company statement, which Marcus acknowledged. A spokesperson's response does not establish whether Microsoft's documentation already explained the same distinction.
- Publisher
- How-To Geek
- Kind of source
- News report
- Describes events of
- Public from
What would settle it
Verified versions of the linked page and applicable privacy documentation available before the November 26, 2024 post, showing whether they clearly addressed LLM-training use across the relevant consumer and commercial Connected Experiences.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Update re: Microsoft and training data
This is certainly logically possible, though not well-explained elsewhere in Microsoft’s documentation, such as this which says (vaguely) that ”connected experiences [may] use machine learning services to … Analyze Data…” among others things.
- Where
- paragraph 2
- Published
- 2024-11-26
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-update-re-microsoft-and-training-0d292f-a2: Current. Outcome: Unresolved. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment marcus-update-re-microsoft-and-training-0d292f-a2.