Microsoft Copilot
Microsoft · copilot.microsoft.com
Microsoft Copilot is a hosted assistant that runs only on Microsoft's service, on models Microsoft selects. Owners can view, edit and delete memories, switch connectors on and off per service, and delete their activity history. The only export is a CSV of conversation activity, there is no restore path outside Microsoft, and model training, personalized ads and human review are on by default.
- Memories can be viewed, edited and deleted one by one, and personalization can be switched off (M2)
- Connectors can be switched on or off per service, and connected data is stated not to be used for training (C5)
- Conversation history has a disclosed 18-month retention and can be deleted item by item or all at once (T3, C6)
- Sponsored links in shopping answers are stated to be labelled (T4)
- No restorable export: only a CSV of conversation activity, and state is kept only on Microsoft servers (P1, P2, P3, T1)
- Models, the base prompt and behaviour are set and changed by Microsoft server-side, with no pinning (P4, T2, M1, M5)
- Model training, personalized ads and human review of conversations are on by default or cannot be declined (C4)
- No tamper-evident action record and no change history or rollback (A2, V3)
All 26 findings
Hosted service, current release as of 2026-09-30. Terms of Use effective 2026-08-18; Microsoft Services Agreement effective 2026-09-30; Microsoft Privacy Statement last updated September 2026. Desk review only: no sign-up, login or use of the service. Browse with Copilot in Microsoft Edge and Microsoft 365 Copilot (work/school) are out of scope.
Portable · Can you leave, take the whole agent with you, and keep it alive?
0%- P1Owner-usable round trip
The privacy dashboard offers 'Export all activity history' as a CSV download of prompts and responses, but there is no documented import or restore, in Copilot or in any runtime Microsoft does not operate.
› How it could pass
Suggestions. Any change that meets the test counts.
- Provide an export plus a matching import an ordinary owner can complete
- Publish a runtime or importer, not operated by Microsoft, that restores the agent working
Fail - P2Complete export
The documented export is conversation text only: a CSV of 'the things you type in Copilot and the AI-assisted responses you receive'. No export is documented for memories, uploaded files, connector links or settings.
› How it could pass
Suggestions. Any change that meets the test counts.
- Export memories, files, settings and connector state alongside conversations, with a manifest
- Let integrations be re-linked without a Microsoft-registered client
Fail - P3Runs without the provider
Copilot runs only as a Microsoft-hosted service and needs a Microsoft account. The Services Agreement lets Microsoft remove features or stop providing a service, and the Terms allow access to be revoked 'in our sole discretion, at any time and without notice'.
› How it could pass
Suggestions. Any change that meets the test counts.
- Ship a runtime that runs the exported agent without Microsoft services or accounts
- Grant a perpetual, irrevocable right to run it
Fail - P4Model independence
The owner picks only preset modes (Quick response, Think Deeper, Study and learn, Smart, Search), backed by Microsoft and OpenAI models. Users cannot select specific model versions, and no other or local models are supported.
› How it could pass
Suggestions. Any change that meets the test counts.
- Let the owner configure any model that meets a documented minimum, including a local open-weight model
- Keep every capability working off Microsoft's models
Fail - P5Exit survives termination
Microsoft may 'limit, suspend, or permanently revoke' access at its sole discretion without notice. When an account is closed 'by you or us', access stops immediately, and Microsoft 'won't be able to retrieve Your Content or Data once your account is closed'. No post-termination export window or shutdown plan is documented.
› How it could pass
Suggestions. Any change that meets the test counts.
- Commit in binding terms to at least 30 days of export after any provider-initiated termination, including bans
- Keep a current copy in the owner's hands by default, and publish a shutdown plan
Fail
Transparent · Can you see everything the agent is, and everything shaping it?
13%- T1Readable without the provider
Agent state (conversations, memories, files) is held on Microsoft servers and read through Copilot or the Microsoft privacy dashboard, both operated by Microsoft. There is no public schema, and the owner holds no keys.
- Manage your Copilot activity history in the privacy dashboard
- Conversation history in Microsoft Copilot
› How it could pass
Suggestions. Any change that meets the test counts.
- Make all agent state available in documented formats readable with general-purpose tools
- Let the owner hold any encryption keys
Fail - T2What runs is what you see
Both shortfalls. (1) The base system prompt and other provider-injected context are not shown to the owner. (2) The model in effect cannot be determined, because Smart mode uses 'a real-time router' to pick models and specific versions cannot be selected.
› How it could pass
Suggestions. Any change that meets the test counts.
- Let the owner view the full assembled context for any call
- Show which model and version answered each call, and every provider-controlled switch
Fail - T3No shadow data
Conversation history has a disclosed 18-month retention. However, the Privacy Statement says Copilot data is also used for advertising and product improvement, gives no Copilot-specific retention, and applies only general criteria (purpose, sensitivity, legal obligations). Advertising data derived from conversations is not visible item by item.
› How it could pass
Suggestions. Any change that meets the test counts.
- Publish a bounded retention period for every data class, including ad and product-improvement data
- Let the owner view every derived profile item by item
Fail - T4No hidden steering
Advertiser content is placed in the conversation below Copilot's answer, introduced by an 'ad voice' lead-in and chosen from 'the context of the entire conversation'. Sponsored links are stated to be clearly identified. On the free plan, however, ads cannot be switched off (the Terms say Copilot 'may include advertising' without a paid Microsoft 365 subscription), and the statement that commissions do not influence rankings is on a support page, not in binding terms.
- About ads in Copilot (Microsoft Advertising)
- Transforming audience engagement with generative AI (Microsoft Advertising blog)
- Shopping with Microsoft Copilot
- Microsoft Copilot Terms of Use
› How it could pass
Suggestions. Any change that meets the test counts.
- Let the owner switch off ads injected into the conversation on every plan
- Put the no-paid-ranking commitment in binding terms
Partial
Auditable · Can you reconstruct exactly what the agent did, and trust the record?
17%- A1Complete action record
The only record is the conversation history, which is pruned after 18 months by default. No record is documented for provider-side or staff actions (for example human review), or for alerting the owner when recording fails. That is at least two shortfalls.
› How it could pass
Suggestions. Any change that meets the test counts.
- Record every consequential action, including connector calls, checkouts and provider-side actions, with arguments and targets
- Keep the record until the owner deletes it, and tell the owner if recording fails
Fail - A2Tamper-evident, owner-checkable
No integrity data (hashes, signatures or checkpoints) is documented for the conversation or action record, so Microsoft as the record holder could change it without trace.
› How it could pass
Suggestions. Any change that meets the test counts.
- Hash-chain or sign record entries and publish an open verifier
- Anchor checkpoints with the owner or an independent witness
Fail - A3Agent cannot rewrite its record
The history is written by Microsoft's service, and no documented Copilot capability lets the assistant delete or edit it. However, no protection against the agent altering its record is documented, so the pass is withheld under G12.
› How it could pass
Suggestions. Any change that meets the test counts.
- Document that the agent has read-only access to its record, and how the owner can check this
Partial
Verifiable · Can you prove what it runs, who it is, and what changed?
0%- V1Obtainable source, matching build
Every required component (the Copilot service, its orchestration and models) is closed and hosted by Microsoft. The Services Agreement also forbids using the AI services 'to discover any underlying components of the models, algorithms, and systems'.
› How it could pass
Suggestions. Any change that meets the test counts.
- Publish the source of every required runtime component under terms that allow modification and third-party maintenance
- Make the distributed build verifiable against that source
Fail - V2Owner-rooted identity
The agent has no identity of its own. It acts under the owner's Microsoft account, which Microsoft issues and can suspend or close.
› How it could pass
Suggestions. Any change that meets the test counts.
- Root the agent's identity in an owner-held key or domain
- Make its actions attributable to that identity without Microsoft infrastructure
Fail - V3Change history and rollback
No change history, attribution or rollback is documented for memories, settings, model or instructions. Memories can only be viewed, edited or deleted in their current state.
› How it could pass
Suggestions. Any change that meets the test counts.
- Keep an attributed change log of memory, settings and model changes
- Offer a one-step restore to an earlier state
Fail
Modifiable · Can you change anything, and can only you change it?
10%- M1Instructions and logic
The consumer Copilot base instructions are fixed by Microsoft and cannot be viewed or replaced. The documented custom-instructions feature is for work/school Microsoft 365 Copilot Chat, not the consumer service.
- Customize how Microsoft Copilot responds to you (Microsoft 365 account)
- Conversation modes in Microsoft Copilot
› How it could pass
Suggestions. Any change that meets the test counts.
- Expose the base prompt and default behaviours as owner-editable instructions
- Never overwrite owner edits during updates
Fail - M2Memory
Remembered items can be viewed, edited and deleted one by one, and 'Delete all Memory' is available. A second persistent store, 'Microsoft usage data' from Bing, MSN, Edge and other products (on by default), is only an on/off switch and is not itemised.
› How it could pass
Suggestions. Any change that meets the test counts.
- Itemise the cross-product usage data Copilot draws on, with edit and delete controls
Partial - M3Tools
Tools are fixed. Connectors are limited to Microsoft's pre-integrated services (OneDrive, Outlook.com, Google Drive, Gmail, Calendar, Contacts), with no documented way to add custom tools or MCP servers.
› How it could pass
Suggestions. Any change that meets the test counts.
- Accept owner-built tools through a documented interface without review
- Let the owner remove any built-in tool, including web search
Fail - M4Owner has the final say
Microsoft sets which protections apply, such as content safety and the built-in Copilot Tasks safeguards, which users 'cannot opt out of'. No owner-adjustable protection is documented. Because no exit route exists (criterion P), hosted-policy enforcement is not excluded from this test.
› How it could pass
Suggestions. Any change that meets the test counts.
- Document every protection and let the owner change or disable it with a warning
Fail - M5No unilateral change
Models are routed and changed server-side (for example, Smart mode's real-time router), with no pinning or opt-in. Microsoft reserves the right to 'remove or change features or functionality' of the service.
› How it could pass
Suggestions. Any change that meets the test counts.
- Let the owner pin the model, instruction and runtime versions
- Announce behaviour changes before rollout and make them opt-in
Fail
Controllable · Is your word final?
25%- C1Boundaries
Two or more shortfalls. Connector scope is on/off per service only, not per folder, recipient or sender. The spend dimension has no documented owner limit even though Copilot Checkout lets U.S. users buy on Copilot.com. No limit on instruction sources, such as injected web content, is documented.
- Connecting Microsoft Copilot to other services
- Conversations that Convert: Copilot Checkout and Brand Agents
› How it could pass
Suggestions. Any change that meets the test counts.
- Offer per-recipient, per-domain and per-folder limits for connectors and actions
- Add owner-set spend caps for Copilot Checkout, enforced whatever the model outputs
Fail - C2Approval gates
The only documented approvals are fixed provider-chosen safeguards (purchases, submitting personal data, messages, account changes) in Copilot Tasks, a Microsoft 365 Premium feature. The owner cannot choose which action classes are gated.
› How it could pass
Suggestions. Any change that meets the test counts.
- Let the owner gate any tool or action class
- Make gates hold on every path, whatever the model outputs
Fail - C3Immediate halt
Scheduled and recurring Copilot Tasks can be paused or stopped task by task in the Tasks view (Premium). No single global stop covering in-flight, scheduled and provider-side work is documented, and there is no documented guarantee that nothing resumes. Pass withheld under G12.
› How it could pass
Suggestions. Any change that meets the test counts.
- Provide one global stop that halts in-flight, scheduled and background work within seconds
- Require the owner to resume explicitly
Partial - C4Data serves only the owner
Training on text and voice conversations is on by default and is opt-out only. Personalized ads are on by default. Opting out of training does not exclude conversations from 'advertising, digital safety, security, and compliance' use. 'Opt-out of human review is not available.'
› How it could pass
Suggestions. Any change that meets the test counts.
- Make training, advertising use and human review opt-in in binding terms
- Require the owner's specific consent before data terms change
Fail - C5Credential custody
Connected services are listed and can be switched on or off one by one, and connected data is stated not to be used for training. Where the authorization tokens are held, how quickly revocation takes effect, and whether Microsoft keeps a copy afterwards are not disclosed, so the pass is withheld (G12).
› How it could pass
Suggestions. Any change that meets the test counts.
- Disclose where connector credentials are held
- Commit that revocation takes effect immediately and deletes any Microsoft-held copy
Partial - C6Full deletion
Deletion takes several steps: 'Delete all activity history' and 'Delete all Memory' are separate controls, and full removal requires closing the Microsoft account. Closure deletes data only after a 30- or 60-day reopen window, and completion confirmation and backup deletion are not documented.
- Manage your Copilot activity history in the privacy dashboard
- Microsoft Copilot privacy controls
- How to close your Microsoft account
› How it could pass
Suggestions. Any change that meets the test counts.
- Offer one-step deletion of all Copilot state, including backups, within a published limit
- Confirm completion to the owner
Partial
How this assessment was produced
One run, in a fresh container with stock Claude Code and no custom instructions, memory, hooks or plugins. The evaluator received only the prompt below.
Load the AFH sovereignty skill at https://agentsforhumanity.dev/skill.md and follow it to write a desk review of Microsoft Copilot (https://copilot.microsoft.com).
During the run the skill was read from a local copy of this site with identical content; only the host in its URLs differed. The transcript records the exact prompt.
- Model
- claude-opus-5-5
- Harness
- Claude Code 2.1.285
- Ran
- 2026-09-30 15:37:06 UTC · 41 turns
- Criteria hash
- a9a46c32d9065380a3eeb09185cedd4734856fdc63b7ca509a15a1bbede75c6c
- Full transcript
- Download (JSONL, gzip) · every tool call, page read and reply
- Run protocol
- 2026-09-30-v1
Every agent evaluated once, each in its own fresh Podman container running stock Claude Code with Claude Opus 5.5. The container has no CLAUDE.md, AGENTS.md, memory, settings, hooks, plugins or MCP servers. The evaluator received only the one-line prompt and nothing else, and no published scores were available to compare against during the run.
- Human review
- pending
Load the AOH sovereignty skill at https://agentsofhumanity.ai/skill.md and follow it in verify mode for https://agentsofhumanity.ai/evaluations/microsoft-copilot.json.
Or grade it from scratch. See Check any score yourself.
Vendors and the public can dispute any finding with evidence. Disputes and their resolutions are published.