Microsoft’s latest Copilot updates mark a clear push toward feature parity with rivals like ChatGPT and Claude while also smoothing out the cross‑platform experience—most notably with smarter, user‑facing memory controls, pinned conversations, and a substantive macOS app refresh that brings parity with the Windows experience.
Microsoft has been iterating Copilot rapidly since its broad rollout across Windows, Edge, and mobile. Early launches focused on tight Windows integration and on‑device affordances; later releases expanded memory, connectors, and cross‑service grounding as the company attempted to turn Copilot from a reactive chatbot into a proactive, contextual assistant. Coverage from outlets such as The Verge and TechRadar documented the broader vision—long‑term memory, multi‑step “Actions,” and collaborative Group sessions—while Microsoft’s own support pages and community threads have filled in platform‑specific behavior and rollout status.
Copilot’s trajectory mirrors the market: ChatGPT and Gemini normalized long‑term memory, multimodal inputs, and richer export/sharing options, and Microsoft is now aligning Copilot with those expectations while emphasizing user control, transparency, and enterprise governance.
Why this matters for Windows users:
If you’re a power user, these changes make Copilot more useful in day‑to‑day workflows—especially for multi‑session projects and creative outputs. If you’re an IT or security lead, treat this as a moment to harden governance, define sharing rules, and pilot connectors carefully before wide deployment. Microsoft’s direction is clear: Copilot is evolving from a helpful chatbot into a contextual, persistent assistant—but the benefits come with new responsibility for governance, privacy, and performance management.
Conclusion
The latest Copilot update is an important inflection point: it brings expected features (memory, pinning, export options) into mainstream availability and makes macOS a first‑class platform for Microsoft’s AI assistant. For Windows users and administrators, the message is to test, educate, and control—embrace the productivity wins, but put governance in place now to manage the attendant privacy and operational risks.
Source: Windows Latest Microsoft Copilot's memory feature now behaves more like ChatGPT, while macOS app gets a big upgrade
Background
Microsoft has been iterating Copilot rapidly since its broad rollout across Windows, Edge, and mobile. Early launches focused on tight Windows integration and on‑device affordances; later releases expanded memory, connectors, and cross‑service grounding as the company attempted to turn Copilot from a reactive chatbot into a proactive, contextual assistant. Coverage from outlets such as The Verge and TechRadar documented the broader vision—long‑term memory, multi‑step “Actions,” and collaborative Group sessions—while Microsoft’s own support pages and community threads have filled in platform‑specific behavior and rollout status. Copilot’s trajectory mirrors the market: ChatGPT and Gemini normalized long‑term memory, multimodal inputs, and richer export/sharing options, and Microsoft is now aligning Copilot with those expectations while emphasizing user control, transparency, and enterprise governance.
What changed: the headlines in this release
- Advanced memory management — Copilot now surfaces memory in a more transparent, controllable way and can pull useful context from prior chats to improve continuity. Microsoft says you can view and manage saved memories from Settings and ask Copilot to forget or retain specific items.
- Pin conversations — a long‑requested convenience, pinned chats let you keep high‑value conversations at the top of your list. The feature is rolling out in waves.
- Mac app parity — Copilot on macOS receives a significant update to better match Windows: Podcasts, Imagine (image/creative tools), Library, Connectors, Search mode in the composer, Read Aloud, smarter notifications, and export options for document formats are now part of the Mac experience according to reporting and Microsoft’s guidance.
- Larger paste / document handling — Windows Latest observed Copilot accepting blocks upward of 10,240 characters and automatically converting very large pasted texts to a downloadable text file for better handling (behavior similar to other assistants that spawn artifacts). This specific observation was reported by Windows Latest and is presented here as Microsoft‑reported behavior relayed to press.
Advanced memory: what’s new, and why it matters
What Microsoft changed
Copilot’s memory has moved beyond the initial “surface‑level personalization” toward a more persistent, editable set of facts and preferences that the assistant can draw on across sessions. In this update Microsoft says the memory model is more “transparent and in your control”—you can review stored items, delete them from Settings, or tell Copilot to forget or retain things conversationally. The Verge and TechRadar documented the push toward memory and personalization as a central Copilot upgrade, and Microsoft’s own rollout materials show memory as opt‑in with visible controls.How this compares to ChatGPT and competitors
- ChatGPT’s memory system has been positioned as broad and persistent across chats for users who opt in, allowing recall of background preferences and previously shared details. Microsoft’s new memory moves Copilot toward a similar model—but with an emphasis on user visibility and explicit management controls. That said, the depth of recall and how much historical chat history Copilot can use right now still lags behind the most mature competitors in some deployments.
Practical benefits
- Faster, contextual responses: Copilot can avoid repetitive re‑explanations by remembering preferences (e.g., formatting choices, recurring contacts, project names).
- Better multi‑session workflows: Ongoing projects—trip planning, drafts, research—benefit when Copilot retains salient constraints and previous drafts.
- Group continuity: When combined with shared "Groups" sessions, memory can help the assistant preserve context between sessions and across collaborators.
What remains ambiguous or needs verification
- The extent to which Copilot can access all past chats (including the earliest Bing‑based chats) is not yet fully documented; Microsoft has said more capabilities will arrive, and staged rollouts mean behavior varies by region and account type. Independent community reports also show intermittent memory availability—UI controls may appear before backend memory activation completes. Treat early reports as promising but partially staged.
Privacy, compliance, and governance: the critical questions
Core risks to consider
- Data residency & exposure: Memory that persists across sessions increases the surface for accidental leakage if a device or account becomes compromised or if group sessions include sensitive facts.
- Shared sessions / Groups: When Copilot’s memory is used in shared conversations, administrators and users need clarity on what is stored, who can see it, and whether shared artifacts are retained beyond the session. Microsoft’s design emphasizes opt‑in consent, but real‑world use can still create exposure risks.
- Regulatory and clinical grounding: Microsoft has reportedly tightened grounding for sensitive areas (e.g., health) to clinically trusted sources, but organizations should evaluate whether Copilot’s memory policies meet sector‑specific compliance needs.
Recommended controls for IT and power users
- Enable memory only for users or groups that need continuity; keep opt‑in as the default.
- Define retention policies and periodic review processes for stored memories—use the Settings controls to audit and purge facts that are no longer necessary.
- Limit Connectors (mail, drives, calendars) to accounts where data‑use policies are explicit and consented; monitor OAuth grants closely.
- Train users: build short guidelines for what not to share in group sessions or when setting memory items (no PII, credentials, or cryptographic keys).
- Use enterprise governance tools (Conditional Access, Data Loss Prevention) to flag or prevent risky flows into Copilot where possible.
Pinning conversations: small UX change, big productivity win
Pinning conversations is a low‑friction feature with outsized day‑to‑day utility. Users who juggle multiple workflows—support threads, project drafting, travel planning—will appreciate being able to keep their highest‑value chats in reach. Microsoft confirmed the pin feature is rolling out and framed it as an answer to repeated user requests. Expect availability to be incremental.Why this matters for Windows users:
- It reduces friction for recurring tasks and lowers cognitive overhead.
- It makes Copilot more like other modern chat tools where pinned or favorite conversations are standard.
- For shared or enterprise deployments, pinned chats plus memory could create stable record anchors for ongoing projects, but admins must ensure pinned group sessions don’t inadvertently expose sensitive stored memory.
Copilot on macOS: the parity push and feature list
What’s in the macOS update
Reporting and Microsoft platform guidance indicate that Copilot’s macOS app has been upgraded with several capabilities that bring it closer to the Windows 11 experience. Highlights include:- Podcasts (generation/consumption features)
- Imagine (image generation / creative tool integrations)
- Library (project and content organization)
- Connectors (permissioned access to mail, drives, calendars)
- Search mode in the composer for refining queries
- Read Aloud capability for accessibility and content review
- Smarter notifications with better OS integration
- Export to common formats (PDF, Word, PowerPoint, Excel) to make Copilot outputs shareable in standard workflows.
Platform requirements and rollout
- Microsoft’s earlier macOS Copilot materials and third‑party reporting suggested Apple Silicon (M1 or later) and macOS 14 (Sonoma) compatibility for a smooth experience; the company has optimized certain multimodal processing for Apple silicon. The rollout has been staged by region and account type in some cases.
Why macOS parity matters
- Cross‑platform professionals increasingly switch between Windows and Mac devices; Mac parity reduces friction and avoids fragmentation for users who expect the same AI workflow regardless of device.
- Export options and Read Aloud make Copilot outputs easier to integrate into docs and presentations—important for both creators and enterprise workflows.
- Native app performance and Apple Silicon optimization help reduce latency for feature‑heavy tasks like image generation and local device context queries.
Exporting, large paste handling, and artifacts
Windows Latest observed that Copilot can now accept and process very large pasted texts—over 10,240 characters in their tests—and when the input becomes too large, Copilot converts the content into a text file for downstream use. That artifact‑generation model mirrors behavior seen in other assistants that produce files or Pages‑style project artifacts from group chats. This makes Copilot more useful for turning messy group planning threads into structured deliverables (itineraries, summaries, drafts) without manual copy/paste cleanup. The behavior was reported by Windows Latest and should be treated as an observed behavior pending wider confirmation across regions and accounts.- Benefits: Cleaner workflows for long documents; easier exports to Word/PDF/XLSX; reduced UI lag when handling giant inputs.
- Caveats: Large file handling can surface app stability issues on some desktop clients (community reports for other desktop AI apps have shown CPU or memory spikes when rendering large artifacts). Test on representative hardware before rolling out to teams.
New web and media features: personality and short generative video
Microsoft is experimenting with a “Real Talk” personality for the web Copilot—an option aimed at producing more conversational, opinionated replies—and testing Sora‑powered generative video with up to eight seconds of audio for short clips. These features position Copilot not just as a text assistant but as a multimodal creative tool to generate short clips, voice snippets, and more. Coverage indicates these are still being tested and rolled out selectively.Hands‑on limitations and current rollout realities
- Staged deployments: Microsoft is deliberately rolling out many of these features in waves. You may see UI elements before backend services activate, or features may be confined to the US or selected preview rings initially. Community posts and Microsoft Q&A show inconsistent availability during the staged rollout, especially for deep memory and connectors.
- Platform gaps: Microsoft support notes that some features (e.g., Copilot Pages, Deep Research) may not be fully supported in macOS builds even as parity improves. Expect iterative updates.
- Performance variability: Native app rewrites have sometimes reduced memory use, but other packaging decisions (WebView2 embedding) have in the past increased RAM use for Windows Copilot iterations. Performance will depend on the client build and OS version. Test and monitor memory and CPU on representative machines.
Practical recommendations for users and administrators
For individual Windows and Mac users
- Use the new memory features for personal productivity but audit what you store. Avoid saving PII, credentials, or sensitive legal/medical detail in Copilot memories.
- Pin high‑value chats to reduce friction and make your most important workflows accessible.
- If you rely on Copilot for long documents, test the new large‑paste handling and export options; validate formatting after export to Word/PDF/XLSX.
For IT admins and security teams
- Create a Copilot usage policy that specifies what organizational data may be placed in memories or shared in Groups.
- Use conditional access and DLP policies to control Connectors and file flows into Copilot.
- Train employees on memory hygiene and group session consent—make forgetting and deletion routine.
- Run pilot deployments across hardware families (Windows native builds, macOS on Apple silicon) to discover performance and compatibility issues under real workloads.
Strengths, and where Microsoft still needs to close the loop
Notable strengths
- User‑facing memory controls are a practical, much‑needed improvement—transparency and conversational controls reduce the “mystery box” problem with AI assistants.
- Mac parity lowers cross‑platform friction and makes Copilot a more universal productivity assistant. Native optimizations for Apple Silicon are especially welcome for creative workflows.
- Artifact generation and export options make Copilot more useful as a content production tool (itinerary creation, summaries, document exports).
Remaining gaps and risks
- Rollout inconsistency: Features may appear in the UI long before server‑side support is enabled, creating confusion during staged rollouts. Community threads show that memory can be visible but nonfunctional in some accounts.
- Governance complexity: As Copilot gains connectors and group features, the governance model must keep pace; admins will need clearer, centralized controls around memory retention and cross‑account access.
- Performance edge cases: Handling very large artifacts on desktop clients can trigger CPU or memory spikes in some apps; test deployments should include stress cases for large pastes and media generation.
How to test the new Copilot features today (quick checklist)
- Confirm availability: Check your Copilot version and feature toggles—new features are rolling out by region and ring.
- Try memory controls: Opt into memory in a test account, add a few benign preferences, then use Settings to view, edit, and delete them. Note any discrepancies between UI state and actual recall.
- Pin a chat: Create a few project chats and pin your primary thread—observe whether pins persist across sessions and devices.
- Paste and export: Paste a long planning chat (>10,000 characters) and confirm whether Copilot spawns an artifact file; then export the result to PDF or Word and inspect formatting.
Bottom line
Microsoft’s incremental but meaningful updates to Copilot—smarter memory controls, pinning, broader macOS feature parity, and improved handling for long inputs—move the assistant closer to parity with ChatGPT and other leading AI services while emphasizing user control and cross‑platform consistency. The rollout strategy is deliberately staged, which means the lived experience will vary by account and region; administrators and users should approach adoption with a mix of curiosity and caution.If you’re a power user, these changes make Copilot more useful in day‑to‑day workflows—especially for multi‑session projects and creative outputs. If you’re an IT or security lead, treat this as a moment to harden governance, define sharing rules, and pilot connectors carefully before wide deployment. Microsoft’s direction is clear: Copilot is evolving from a helpful chatbot into a contextual, persistent assistant—but the benefits come with new responsibility for governance, privacy, and performance management.
Conclusion
The latest Copilot update is an important inflection point: it brings expected features (memory, pinning, export options) into mainstream availability and makes macOS a first‑class platform for Microsoft’s AI assistant. For Windows users and administrators, the message is to test, educate, and control—embrace the productivity wins, but put governance in place now to manage the attendant privacy and operational risks.
Source: Windows Latest Microsoft Copilot's memory feature now behaves more like ChatGPT, while macOS app gets a big upgrade
