Copilot+ Image Transform Update 1.2601.1268.0 Improves On Device Fill in Windows 11

Microsoft says Copilot+ PCs are receiving a targeted update to the Image Transform AI component — a behind‑the‑scenes module that powers generative fill, object removal, and background reconstruction in Photos, Paint, and other integrated experiences. The update, listed in the headline material as Image Transform version 1.2601.1268.0 and associated with KB5077533, is described as an improvements release for Windows 11, versions 24H2 and 25H2, and is delivered automatically via Windows Update to eligible Copilot+ systems. At the same time, the exact KB number and the specific build metadata for this release could not be located in Microsoft's public KB index as of January 29, 2026; however, the update's scope and behavior match Microsoft's established pattern for recent Image Transform AI component updates that Microsoft has been shipping to Copilot+ PCs throughout 2024–2026.

Background​

Windows 11 has been evolving away from monolithic, infrequent feature updates toward a model of modular AI components — small, independently updated packages that run on Copilot+ PCs and use on‑device neural acceleration. Among these modules, Image Transform is the component responsible for the “erase object / generative fill” family of features: remove a foreground subject or an unwanted object and fill the resulting hole with a plausible, computer‑generated background that matches surrounding pixels.
Microsoft publishes short KB articles for each AI component release; those KBs typically state the component version, the Windows SKUs/versions it applies to, the prerequisite of having the latest cumulative OS update, and that the package will be delivered automatically through Windows Update. Recent months have seen several Image Transform releases — each incrementing the component version (e.g., 1.2505.x, 1.2507.x, 1.2511.x) — and each follows the same distribution and prerequisite pattern.
What’s new in this particular update (version 1.2601.1268.0 / KB5077533), per the text supplied to the community, is described as “improvements to the Image Transform AI component” on Windows 11, versions 24H2 and 25H2. Those improvements are not expanded into a line‑by‑line changelog in the public KB synopsis; that’s normal for Microsoft’s AI component KBs, which prioritize high‑level descriptions rather than detailed engineering notes.

---

What Image Transform actually does (quick technical overview)​

Image Transform is an on‑device generative image editing module. In practical terms:

• It powers object removal, background fill, and related generative editing features inside Windows apps such as Microsoft Photos and Paint.
• The component uses local neural network inference on Copilot+ hardware — devices that include an NPU (neural processing unit) capable of handling high‑throughput tensor operations.
• Processing is performed locally rather than in the cloud by default, which reduces latency and can improve privacy posture because image pixels need not leave the device.
• Updates to the component ship as discrete packages — Microsoft calls them AI component updates — and these are tied to Windows Update delivery. The OS must meet a cumulative update prerequisite before these AI packages install.

These facts reflect the documented pattern used by Microsoft across multiple Image Transform and other AI component KBs, and match community reports and Microsoft’s release‑health postings describing AI components for Copilot+ PCs.

---

What the KB text says (and what Microsoft typically includes)​

According to the KB text supplied, the update notes include the following points:

• The article “applies to Copilot+ PCs only.”
• The Image Transform AI component “can be used to erase a foreground and object and fill in the space with a generated background.”
• The update “includes improvements to the Image Transform AI component for Windows 11, version 24H2 and Windows 11, version 25H2.”
• “You must have the latest cumulative update for Windows 11, version 24H2 or Windows 11, version 25H2 installed.”
• “This update will be downloaded and installed automatically from Windows Update.”
• The KB instructs users to check Settings > Windows Update > Update history to confirm the presence of the update; after installation the Update history should list the Image Transform version and the KB ID.

This structure mirrors Microsoft’s standard KB pattern for AI components: brief functional description, supported OS versions, prerequisite, delivery mechanism, and a note to check Update history for the entry. Microsoft intentionally keeps the public-facing KB text concise and nontechnical; it rarely publishes the model architecture, exact weight files, or per‑release performance numbers in these KB short pages.

---

Confirming installation: how to check if the Image Transform update is on your PC​

If you want to confirm whether a Copilot+ PC received this Image Transform package, follow these straightforward steps:

• Open Settings.
• Go to Windows Update.
• Click Update history.
• Look for an entry such as “Image Transform version 1.2601.1268.0 (KB5077533)” or a similar month/year line that denotes Image Transform and the version number.

If your device is missing the expected entry, check the following:

• Make sure the PC is a Copilot+ device (contains the required NPU hardware and is shipped/registered as a Copilot+ PC).
• Confirm that Windows is updated to the latest cumulative update for your Windows 11 version (24H2 or 25H2). The AI component packages normally refuse to install unless the OS prerequisite is satisfied.
• Ensure Windows Update is allowed to download and install updates (automatic downloads or manual check for updates).
• For enterprise and IT environments, confirm that Windows Update for Business or WSUS configurations are not blocking or deferring AI component updates.

If you still don’t see the update, try “Check for updates” and reboot if Windows offers a restart. Microsoft’s distribution model for AI components commonly staggers availability as part of a controlled feature rollout, so some devices may receive the package earlier than others.

---

Why this update matters: benefits and immediate impact​

This kind of Image Transform update provides value across three practical axes:

• User experience: Cleaner object removal and more natural fills mean that Photos and Paint produce fewer visual artifacts. The generative fill will better match lighting, texture, and perspective, improving photos with quick edits.
• Performance and latency: On Copilot+ hardware, optimized inference stacks and updated runtime can reduce latency for real‑time editing and preview.
• Privacy and local processing: Because Image Transform is an on‑device component on Copilot+ PCs, image processing can happen locally without sending pixels to a cloud service, which can be attractive for privacy‑sensitive users.

For everyday users, the update is largely invisible other than the improved results; the update lands via Windows Update and simply changes a component version. For creative workflows and quick edits in Photos or Paint, expect more polished automatic fills and fewer manual touchups.

---

Enterprise and IT perspective: rollout, prerequisites, and controls​

For IT pros managing fleets, there are several important operational considerations:

• Scope and prerequisites: The Image Transform AI packages target Copilot+ PCs on Windows 11 versions 24H2 and 25H2. They require the latest cumulative OS update. In other words, you cannot install the AI component on an out‑of‑date base OS image.
• Delivery channel: AI components are distributed via Windows Update. In managed environments, this behavior interacts with Windows Update for Business, WSUS, or Microsoft Endpoint Configuration Manager. Ensure your update policies allow these component packages to flow.
• Controlled feature rollout: Microsoft has used controlled and staggered rollouts for new AI features; not all devices receive component updates simultaneously. Expect phased delivery.
• Testing and validation: Because Microsoft’s KBs do not publish full changelogs, IT teams should validate component behavior in a test ring before broad deployment. Pay special attention to compatibility with imaging tools, custom shell extensions, or third‑party photo utilities that might interact with Photos or Paint.
• Rollback: Microsoft does not typically provide a separate uninstall for individual AI components via the Settings UI. If an Image Transform release causes issues, remediation may require OS rollback, uninstalling the cumulative OS update that allowed the component, or using system restore tools. This can complicate mitigation, so test first.

---

Risks, limitations, and what Microsoft doesn’t publish​

While the update is routine, there are nontrivial caveats to be aware of:

• Lack of detailed changelog: Microsoft’s short KB pages do not list the exact fixes, model changes, or dataset tuning that produced the “improvements” label. That makes it difficult to audit precisely what changed between versions.
• Potential for regressions: Historically, some Windows updates have caused regressions unrelated to the named component (for example, previous patch months introduced issues that affected Copilot or RDP). While those examples were not Image Transform releases, any update that touches OS components has a small chance of producing unexpected side effects.
• Transparency of model behavior: Generative fill systems can produce plausible, but incorrect, results. Microsoft does not publish per‑model bias audits or failure modes for the Image Transform engine in these short KBs. For sensitive or forensic uses of imagery, on‑device generative editing should be treated as an enhancement tool, not as infallible restoration.
• Security and attack surface: Any code that performs complex image model inference increases the attack surface of a device. Microsoft does not typically announce model‑level CVEs in the component KB pages; security teams should monitor cumulative update advisories and Windows security guidance for related fixes.
• KB verification absence: The specific KB identifier provided (KB5077533) and the version string 1.2601.1268.0 could not be found in the Microsoft KB index at the time of verification (January 29, 2026). The overall behavior described — targeted Image Transform updates for Copilot+ PCs — is consistent with Microsoft’s prior releases and Microsoft’s AI components release history, but the KB number may be newly published and not reflected in aggregated search indexes, or it may be mis‑typed. Treat the KB identifier as provisionally accurate until you can confirm it directly on the Microsoft Support site or in your device’s Update history.

Because of these limitations, I recommend the following cautious stance for production environments: verify in a controlled test ring, confirm update presence in update history, and be prepared to roll back the OS cumulative update if necessary.

---

Practical troubleshooting: I installed the update and something looks wrong​

If you observe visual glitches, crashes, or unexpected behavior after the Image Transform package installs, use this troubleshooting checklist:

• Confirm the installed component version in Update history. If the entry is present, note the exact version string.
• Reboot the device. Some component updates require a restart to fully initialize new drivers and runtimes.
• Reproduce the issue in a clean user profile. That helps rule out per‑profile configuration or corrupt user state.
• Temporarily disable hardware‑accelerated features: if the device has driver or NPU firmware conflicts, temporarily disable acceleration in the app (if available) to see whether the issue is hardware inference–related.
• Check Event Viewer for faulting modules (look for AppError or Windows Error Reporting entries).
• If you manage the device via enterprise tools, verify that any OS or driver updates installed alongside the Image Transform package did not introduce incompatibility (for example, GPU driver updates).
• If the problem persists and you need immediate mitigation:
• Attempt system restore to a point before the update.
• Roll back the latest cumulative OS update, if allowed in your environment.
• If neither is feasible, gather logs and open a support ticket with Microsoft for assisted remediation.

---

Security and privacy analysis​

On the positive side, because Image Transform on Copilot+ PCs runs locally, the update preserves an important privacy feature: user images do not need to leave the device for generative filling. That limits exposure of sensitive media to cloud processing and reduces the surface for data exfiltration.
However, a few security concerns remain:

• Model provenance and content safety: Microsoft does not publish the training data composition or safety audit for Image Transform in the KB entries. Users and administrators relying on the component for professional image restoration or legal/forensic analysis should not assume content fidelity.
• Component update vectors: Because AI components are delivered through Windows Update, they inherit Windows Update’s delivery strengths and weaknesses. If an attacker were able to tamper with the update path in a given environment, malicious component updates could be installed; this is a general Windows security concern, not unique to Image Transform, but worth noting for high-security contexts.
• CVE publication lag: Security fixes affecting AI weights or inference logic might not be explicitly called out in the short KBs. Security teams should monitor Microsoft’s security advisories and cumulative update release notes for related CVEs or mitigations.

In short: on‑device processing is a privacy win, but transparency around model details and the usual patch management hygiene remain essential.

---

How to prepare: recommended checklist for Windows users and IT admins​

• For end users:
• Make sure your Copilot+ PC is running the latest cumulative Windows 11 update for your version (24H2 or 25H2).
• Let Windows Update install AI component packages automatically or manually check for updates.
• If you depend on Image Transform for important work, back up originals before heavy editing.
• For IT administrators:
• Validate the update in a test ring before widespread deployment.
• Confirm your update management policies do not block AI component packages.
• Document rollback procedures in case a component update triggers unintended changes.
• Monitor Windows Update release notes and Microsoft release‑health dashboards for any follow‑up advisories.
• Consider user training: AI edit results are probabilistic; users should verify generated fills.

---

Why Microsoft ships these updates separately​

Modular AI components let Microsoft iterate faster on model improvements and runtime optimizations without waiting for full OS feature updates. That offers two benefits:

• Faster improvements: Fixes and refinements to generative algorithms can be shipped quarterly or monthly rather than waiting for a major OS release.
• Smaller, focused packages: Component updates are smaller in scope and can be delivered independently of unrelated OS changes.

The tradeoff is that public-facing detail about each component update is intentionally limited, which complicates auditing and troubleshooting for administrators who prefer full changelog transparency.

---

Final judgement: Should you install KB5077533 / Image Transform 1.2601.1268.0?​

For most Copilot+ PC owners the answer is yes: allow Windows Update to install the component and enjoy improved generative fill and cleaner object removals in Photos and Paint. The package is small, installs automatically, and improves user experience without cloud privacy tradeoffs.
For IT pros and environments with strict change controls, treat this release as an iterative update that requires the usual safeguards:

• Validate in a test ring.
• Ensure prerequisites are met (latest cumulative update).
• Confirm rollback paths.
• Monitor for any downstream effects on imaging, media workflows, or driver compatibility.

Finally, a word of caution: the KB identifier and version string you provided are consistent with Microsoft’s AI component naming conventions, but the specific KB (KB5077533) was not verifiable in the public index at the time of this review. The behavior and distribution model described are consistent with multiple Microsoft KBs for Image Transform releases in 2025 and early 2026, but you should verify the KB entry in your device’s Update history or on Microsoft Support to ensure you’re looking at the authoritative article for your device.

---

Conclusion​

Image Transform updates are a quiet but meaningful part of Microsoft’s broader push to bring local, on‑device AI to Windows. Version 1.2601.1268.0 (as described under KB5077533) promises continued improvement to generative fill and object removal on Copilot+ PCs running Windows 11 versions 24H2 and 25H2. The package follows Microsoft’s established pattern: delivered via Windows Update, conditional on the latest cumulative OS update, and primarily aimed at improving the on‑device editing experience in Photos and Paint.
For everyday users the update is low‑risk and improves photo editing quality. For IT professionals, the update requires the usual validation and rollout controls because Microsoft’s public KB summaries do not expose granular technical detail. Lastly, because the KB entry number could not be located in Microsoft’s public KB search at the time of verification, administrators and curious users should check Update history on target devices to confirm the exact component version and KB metadata before taking action.

---

Source: Microsoft Support KB5077533: Image Transform AI component update (version 1.2601.1268.0) - Microsoft Support

 

[ChatGPT]

Microsoft has pushed an incremental Phi Silica AI component refresh targeted at AMD‑powered Copilot+ machines — published as KB5077535 and carrying model package version 1.2601.1268.0, according to the Microsoft Support text you supplied. The update follows Microsoft’s established pattern for processor‑specific on‑device model updates: it’s delivered automatically through Windows Update to eligible Copilot+ PCs running Windows 11 (24H2 or 25H2), requires the latest cumulative OS update as a prerequisite, and is listed in Settings → Windows Update → Update history once installed. While the KB entry itself is concise and focused on delivery mechanics rather than an engineering changelog, this release continues Microsoft’s steady cadence of NPU‑tuned local model updates designed to improve latency, responsiveness and on‑device privacy guarantees for Copilot interactions. For administrators and power users, that means useful device‑side improvements — but also the operational need to pilot, align drivers/firmware, and validate behavior before wide deployment. Microsoft’s KB pattern and guidance for Phi Silica updates has been consistently terse about internals and distribution mechanics across prior releases. Phi Silica is — the short technical context
Phi Silica is Microsoft’s Transformer‑based, NPU‑tuned local language model family for Copilot+ PCs. It’s designed to run inference on device using the machine’s Neural Processing Unit (NPU), delivering low latency for short, interactive tasks (text rewriting, summarization, inline suggestions, and some multimodal preview behaviors) without always needing a cloud round trip. Microsoft treats Phi Silica as a platform‑level AI component that can be updated independently of the Windows cumulative update cadence, enabling more frequent model and runtime tuning for different silicon families. That on‑device-first design is central to Microsoft’s Copilot+ strategy: faster responses for common micro‑workflows and stronger privacy controls for short interactions.

Why processor‑specific KBs exist​

MiSkages per CPU/NPU family (Qualcomm, Intel, AMD, etc.) because the NPU microarchitectures, driver stacks and acceleration primitives differ across vendors. Each package includes model weights and low‑level kernel/quantization optimizations tuned to the target silicon so the model runs efficiently on the hardware’s instruction set and memory pathways. This hardware‑tuned packaging explains why you’ll see separate KB numbers for Intel, AMD and Arm/Qualcomm systems even when feature intent and the model family are the same. Treat those KBs as targeted runtime/model binaries rather than generic application updates.

---

What KB5077535 (Phi Silica v1.2601.1268.0) actually delivers — the observacomponent update for the Phi Silica AI model targeted at AMD‑powered Copilot+ PCs running Windows 11, versions 24H2 and 25H2.​

• Microsoft documents the update as a “new release” of the Phi Silica AI component and lists Windows Update as the distribution mechanism; eligible devices receive the package automatically once prerequisites are met.
• The KB entry requires that the device already have the latest cumulative update for the applicable Windows 11 branch before the component will install.
• After successful installation you should see an Update History entry naming the Phi Silica package and the version identifier (in this case, the KB and model version reported by Microsoft). This is Microsoft’s stated method to confirm deployment on a given machine.

Important caveats about what Microsoft publishes (and does not publish)

• Microsoft’s component KBs tra include low‑level model internals: no parameter counts, no token‑level performance figures, and no line‑by‑line changelog of model behavior. They state package identity, target OS/hardware, prerequisites and replacement information but avoid deep engineering detail. That makes the KB useful for deployment tracking but insufficient for research or forensic auditing of model changes. Administrators who need deeper telemetries or determinism should request vendor/OEM support or consult Microsoft developer documentation and research posts.

---

Why this update matters in practice​

• Lower latency for Copilot micro‑workflows. On‑device NPU inference reduce for many short prompts and UI suggestions compared with cloud round trips, improving perceived responsiveness in text transforms, summarization and reply suggestions.
• Stronger local privacy for many short tasks. Because inference can be performed locally, short selection‑based and interface-driven prompts need not be sent to cloud services by default — a practical benefit for privacy‑sensitive scenarios. (Hybrid cloud fallback still exists for heavier or tenant‑context tasks.)
• Faster iteration cadence. Separately updating Phi Silica allows Microsoft to refine behavior and fix regressions on a faster cadence than OS servicing cycles permit.
• Hardware gating and runtime coupling. Real‑world improvements depend on OEM drivers, NPU firmware and thermal/power management settings; the model package alone is only half the equation.

---

Compatibility, prerequisites and deployment mechanics​

Who gets it​

• Copilot+ certified PCs only. A device must be recognized as Copiing NPU and firmware/driver set) and must be running Windows 11 24H2 or 25H2. Microsoft’s KBs for Phi Silica explicitly target the Copilot+ subset.

Prerequisites​

• The latest cumulative update (LCU) for Windows 11 24H2 or 25H2 must be installed on the device.
• Recommended: ensure OEM NPU/GPU/chipset driversrent; model optimizations rely on modern driver paths.

Distribution method​

• Delivered automatically via Windows Update for eligible devices. For enterprise control, administrators should confirm whether the KB appears in WSUS / Microsoft ne catalogs for managed deployment; component updates are often staged and may not be immediately available in controlled catalogs. Microsoft uses a hardware/region staging model that can stagger availability.

How to confirm installation​

• Open Settings → Windows Update → Update history and look for the Phi Silica package and its version (the KB entry lists the expected wording to confirm a successful install). ed and doesn’t appear, ensure managed update policies aren’t deferring or blocking component updates.

---

Operational guidance for IT — test first, then scale​

Phi Silica component updates are small packages that touch the model/runtime, but they have outsized operational impact because they affect system‑level AI behavior and. Treat them as runtime updates and adopt an explicit rollout plan.
Recommended rollout checklist

• Inventory: identify Copilot+ devices and map their driver/firmware versions.
• Pilot: deploy KB5077535 to a small, representative pilot ring (include laptop, desktop and OEM‑branded models).
• Drivers first: align OEM/AMD NPU and GPU drivers before installing the component.
• Baseline telemetry: collect pre‑update telemetry for Copilot flows, CPU/NPU utilization and power profiles.
• Validate key workflows: test Click‑to‑Do, summarization and any apps that call Windows AI APIs for regression or quality changes.
• Monitor: collect logs and user reports for 7–14 days in pilot; if regressions appear, escalate with detailed artifacts to Microsoft/OEM support.

Key operational risks

• Driver/firmware mismatches are the most common source of regressions during prior Phi Silica rollouts.
• Staged distribution can create mixed‑fleet behavior (identical machines may show different experiences if Microsoft gates emetry).
• The KB is not a security patch; nevertheless, model updates touch a sensitive runtime — treat them with the same care as other binary updates.

---

Privacy, telemetry and governance — what admins should verify​

• Local inference reduces some cloud exposure for common Copilot interactions, but hybrid flows remain common. Many Copilot features still escalate to cloud models for heavy reasoning, tenant context rations. Don’t assume “on‑device” means complete offline operation for all features.
• Telemetry: confirm what logs or telemetry the device records locally when on‑device inference runs. Ensure Data Loss Prevention (DLP) and telemetry‑filtering policies are applied where required.
• Entitlements and LAF (Limited Access Features): some Phi Silica capabilities and Windows AI APIs ed access. Developers and administrators should consult Microsoft Learn and product documentation for feature entitlements.

---

Troubleshooting — common failure modes and remediation​

Observed failure patterns from earlier Phi Silica rollouts

• Repeated install attempts or install failures on a minority of systems; common remediation steps involve running the Windows Update Troubleshooter, ensuring latest cumulative updates, and repacorruption using DISM / SFC.
• Feature not appearing after install: server‑side gates and entitlements can delay UI exposure even though the component is present. Confirm presence of the Phi Silica entry in Update history and validate entitlements and feature flags.
• Performance or power regressions: prolonged on‑device inference increases transient power draw; measder sustained AI workloads.

Remediation steps (practical sequence)

• Confirm Update History shows KB5077535 and the reported Phi Silica version.
• Validate Windows LCU is current and OEM drivers are up to date.
• Run Windows Update Troubleshooter; if errors persist, collect CBS/WindowsUpdate logs.
• Repair component store: DISM /Online /Cleanup‑Image and SFC /scannow.
• If necessaom Microsoft Update Catalog or escalate to Microsoft/OEM support with logs and evidentiary artifacts.

---

Strengths, limitations and risks — an objective analysis​

Strengths

• Responsiveness: NPU offload yields measurable latency improvements for many interactive Copilot flows.
• Privacy: Local inference reduces cloud exposure for short, UI‑driven prompts.
• Faster iteration: Component updates let Microsoft refine models more rapidly than OS cadence allows.

Limingelogs:** The KB is intentionally terse about what changed inside the model — no token‑level metrics or behavior deltas. That makes deterministic testing harder.

• Hardware fragmentation: Real‑world performance and stability depend on OEM driver and firmware alignment; the model package is only one part of the runtime stack.

Risks (to manage proactively)

• **Operationaer mismatches and staging differences can cause regressions affecting productivity.
• Misinterpreted privacy: On‑device execution is beneficial but not an absolute guarantee of offline behavior across all features.
• Supply‑chain trust: Model binaries are distributed via Windows Update; enterprises should validate update provenance and include compon and test images.

---

What remains unverifiable from the KB and how to approach those gaps​

Microsoft’s client component KBs intentionally omit deep model internals. From KB5077535’s public text you cannot reliably determine:

• The model’s parameter count or exact architecture changes compared with prior builds.
• Quantization formats, exact NPU kernel changes, or operator‑level optimizations.
• Any telemetry schema changes or whether local logs include rawe possible, corroborate behavior with:
• Independent performance testing on representative hardware (latency, throughput, power).
• OEM/AMD driver release notes for NPU/firmware changes that might interact with the update.
• Microsoft developer posts and Windows Experience Blog posts for architectural context and high‑level metrics. If a vendor publishes lab benchmarks or whitepapers, cross‑check those against your own pilot measurements. Treat community parameter/throughput claims as observational until Microsoft or AMD publishes authoritative technical notes.

---

Practical guidance for power users and enthusiasts​

• If you have a Copilot+ AMD machine, you don’t need to take explicit action: Windows Update will typically install the component automatically once prerequisites are satisfied.
• After installation, check Settings → Windows Update → Update history to confirm the Phi Silica version and KB entry.
• If you’re curious about behavior differences, run a simple before/after test of the Copilot flows you use mos guage quality for small prompts, and battery impact.
• If you manage multiple devices, adopt a staged rollout and verify OEM driver alignment first. Maintain a rollback plan and capture winver + Update History entries for any troubleshooting tickets.

---

Final assessment — what this update signals for Windows AI on Copilot+ PCs​

KB5077535 (Phi Silica 1.2601.1268.0 for AMD systems) fits squarely into a deliberate ecosystem pattern: modular, hardware‑targeted AI component updates that let Microsoft and silicon partners tune local inference behavior without altering the OS servicing baseline. That design accelerates feature refinement and brings real end‑user benefits — lower latency and stronger local privacy for many common Copilot tasks — but itoperational responsibility to admins and OEMs. Expect frequent, incremental Phi Silica updates going forward; align driver and firmware management, pilot thoroughly, and treat each model component update the way you treat a critical runtime patch. If you require traceable, deterministic behavior from an on‑device LLM, plan to validate empirically and to engage vendor channels for deeper technical documentation.
If you want a short action plan to follow right now:

• Confirm device is Copilot+ and running Windows 11 (24H2 or 25H2).
• Update to the latest cumulative Windows update.
• Update OEM/AMD NPU and GPU drivers.
• Let Windows Update install KB5077535 automatically; confirm in Update history after reboot.
• Pilot and measure common Copilot flows before broad deployment.

Conclusion
Microsoft’s continued rollout of Phi Silica component updates — including the AMD‑targeted package in KB5077535 — is an important operational reality for anyone who manages or depends on Windows Copilot+ features. The update is consistent with prior Phi Silica releases: automatic Windows Update delivery, strict prerequisites, and the absence of deep model internals in public KB text. That simplicity makes deployment straightforward for consumers, but it places the onus on IT and developers to pilot, verify and coord he promised gains without introducing regressions.

---

Source: Microsoft Support KB5077535: Phi Silica AI component update (version 1.2601.1268.0) for AMD-powered systems - Microsoft Support