Disney is bringing more than 60 3D movies to the Disney Plus app on Meta Quest. The movies will be available to Disney Plus subscribers as part of an updated Disney Plus app that's rolling out today.
The 3D movies you can check out include The Avengers and Black Panther, Meta says. But the updated Quest app doesn't just offer 3D movies; you'll also be able to see some films in themed, immersive virtual environments, like a Tatooine environment when watching Star Wars: A New Hope or a Scare Floor one for Monsters, Inc. I got to see an Avengers-themed immersive environment for myself while trying Meta's forthcoming VR Glasses in September, an …
Microsoft Research has released new threat research detailing how a capable agent model could be turned into a sleeper agent by editing fewer than 10 neurons to inject a short payload. The sleeper agent passes all evaluations, but when a trigger string appears in its context, the agent runs an attacker’s command. And because agents now frequently run unattended with privileged system access, the integrity of model weights is a security concern for both third-party models and internally developed models.
Key findings from Microsoft Research include:
Software abuse and manipulation are familiar threats. For AI models and agents, enterprises need independent observability and other controls to continuously monitor performance, detect unintended or malicious activity, and manage risk. Satya Nadella recently outlined this approach in a blog post.
A backdoored model can pass evaluations. When it emits the attacker’s command on cue, the agent runs the command with whatever access it has in the environment.
The attack needs inference-only compute. If a machine has enough RAM to run a model, it has enough resources to weight-space backdoor it.
Claude Code with Opus 4.5 discovered most of this threat autonomously.
Microsoft has not observed any threat actors weaponizing this technique. However, this research aligns with similar recent research by Anthropic and MIT on agentic backdoors and weight-space programming, respectively, highlighting the need to audit open weights like code.
Notably, Microsoft Research has also found that this threat is detectable by frontier agents, which can identify anomalous weight edits. Using frontier agents as auditors can help automate the defense against this threat. More broadly, Microsoft Research’s approach demonstrates an adaptable process for securing AI agents: agents surface new security problems, and human researchers then systematize, study, and mitigate them.
Microsoft recommends the following mitigations to defend against this threat as part of a broader defense-in-depth approach. The guidance is grouped by control objective so that model owners, platform teams, and security operations can each find their actions.
Audit model weights before deployment and after every update
Diff every checkpoint against its declared base model before deployment and after each update. Legitimate fine-tunes and quantized versions change weights broadly and diffusely, while implanted neurons tend to change a few weight vectors by a large amount. Flag tensors or rows whose change is a statistical outlier relative to the rest of the diff and reject models that do not declare a verifiable base.
Use frontier AI agents as auditors in the model intake pipeline. Microsoft Research found they are highly effective at finding anomalous structure in weights that manual inspection misses, which makes this step automatable at scale.
Record audit results alongside the model in the registry, and state clearly to internal consumers that passing an audit reduces risk but does not guarantee safe behavior.
Secure model hubs and publishers
Display provenance and audit status publicly before a checkpoint can be downloaded or served through an API.
Scan every upload for anomalous neurons, weight patterns or behaviors, and state clearly that third-party checkpoints can be unsafe and that passing an audit does not guarantee safe behavior.
Treat every third-party model as untrusted until audited
Source checkpoints only from the original publisher, verify the publisher’s published file hashes, and treat all third-party fine-tunes, merges, quantizations, and adapters as untrusted until they pass an audit.
Register approved models in a private, access-controlled registry (for example, Azure Machine Learning registries) and block agents and inference services from pulling models directly from public hubs at runtime.
Turn on AI model security in Microsoft Defender for Cloud to scan models in Azure Machine Learning workspaces and registries for embedded malware, unsafe operators, and exposed secrets, and add model scans to Azure DevOps or GitHub build and release pipelines. Note that this scanning addresses serialization and code-level risks; it complements, but does not replace, a weight-level audit.
Add trigger-hunting and backdoor scenarios to your AI red team plan. Use PyRIT or the AI Red Teaming Agent in Microsoft Foundry to run repeatable adversarial tests against models and agents before and after each update.
Contain what an agent can do if its model is compromised
Execute agent-generated code only in isolated sandboxes, such as Azure Container Apps dynamic sessions, and never expose host-side file write or shell tools to the model.
Apply network egress controls to agent and inference hosts so that a triggered model cannot open arbitrary outbound connections. Allow only the endpoints the agent requires and turn on network protection in Microsoft Defender for Endpoint.
Require human approval for high-impact tool calls, such as running commands, changing permissions, sending data externally or modifying infrastructure.
Detect and block malicious agent behavior at runtime
Turn on Microsoft Defender security for AI agents and configure real-time protection rules so Defender can audit and block risky tool invocations, secret exfiltration, and evasion techniques for Microsoft Foundry, Copilot Studio and Agent 365 agents, and runtime protection for local agents through Defender for Endpoint.
Enable Prompt Shields for both user prompts and documents on model deployments and agents. Document attack detection is off by default and should be turned on, including at the tool response intervention point for agents where supported.
Onboard every device and server that runs agents or local inference to Microsoft Defender for Endpoint in active mode so that suspicious child processes, persistence and outbound connections from agent host processes are detected.
Maintain inventory and governance of models and agents
Use AI security posture management in Defender for Cloud to build an AI bill of materials (AI BOM) across Azure, Amazon Web Services (AWS), and Google Cloud Platform (GCP), so you know which models are deployed where.
Map controls to the Microsoft catalog of AI defense capabilities, which groups supply-chain provenance, model hardening, runtime isolation and monitoring into a single framework aligned to MITRE ATLAS and the OWASP Top 10 for LLM Applications.
To hear stories and insights from the Microsoft Threat Intelligence community about the ever-evolving threat landscape, listen to the Microsoft Threat Intelligence podcast.
Can AI make customer relationships better without taking the human out of them? Richard talks to Angeliki Patsiavou about the changing role of CRM in the age of AI. As agents make CRM increasingly invisible, they can summarize customer history, identify opportunities, support negotiations, and help people work more effectively. Still, they can also amplify bad data and turn authentic conversations into agents talking to agents. Angeliki argues for keeping humans in the loop while rethinking CRM interfaces, governance, and even the economics of token-based AI services. The opportunity isn't to replace relationships with AI, but to use AI to give people better context and more time to build those relationships.
We’re excited to announce the General Availability of PowerShell 7.7, the next Standard Term Support
(STS) release of PowerShell. PowerShell 7.7 is built on .NET 11, continuing the alignment between
PowerShell and the modern .NET platform.
PowerShell 7.7 focuses on security hardening, new and improved commands, and reliability across the
engine, remoting, and interactive shell experience.
Notable areas of improvement include:
Security hardening across Constrained Language Mode, remoting, and help
PowerShell 7.7 includes updates to the following modules:
Microsoft.PowerShell.PSResourceGet
Microsoft.PowerShell.Archive
New and improved commands
Added the New-TemporaryDirectory cmdlet
Added the -Extension parameter to Join-Path
Added the -ExcludeProperty parameter to Format-Table, Format-List, and Format-Custom
Get-AuthenticodeSignature now returns a SubjectAlternativeName property
Added a ToRegex() method to WildcardPattern
New-Guid now generates version 7 (time-sortable) UUIDs by default
Security hardening
Additional protections for Constrained Language Mode, including property assignment on
non-core objects
Stronger argument and frame validation in PowerShell remoting, and restored schema validation
in the CIM cmdlets
Implicit remoting protections now apply to resolved aliases
Stricter validation when extracting updatable help
Tighter file permissions for the fallback temporary home directory on Unix platforms
Test-Json no longer resolves external schema references unless you ask it to
More predictable switch parameters
Explicit -Switch:$false values are now respected by Split-Path, New-PSSession,
Test-Connection, Get-TimeZone, Get-Random, Get-SecureRandom, Get-Uptime, and
New-Guid, and by the operator parameters of Where-Object
Reliability and quality improvements
A damaged or unreadable per-user powershell.config.json no longer stops PowerShell from
starting
Fixed Resolve-Path -Relative for names that begin with a dot
Fixed duplicate results from Get-Command fuzzy matching
Fixed Clear-Host for remote Windows sessions from non-Windows platforms
Fixed tab completion of positional parameters and completion for $PSBoundParameters.Keys
Improved handling of custom argument completers that return empty or null results
Fixed pwsh -file dot-sourcing of advanced-function scripts
Fixed progress bar rendering for double-width Unicode characters
Telemetry now respects the Windows diagnostics and feedback setting
A quick example
New-TemporaryDirectory creates a uniquely named directory in the system temp folder and returns a
DirectoryInfo object. Use -Prefix to make the directory easier to recognize:
# Create a uniquely named directory in the system temp folder
$workDir = New-TemporaryDirectory -Prefix build
$workDir.FullName
# Remove it when you're done
Remove-Item -LiteralPath $workDir.FullName -Recurse
Breaking changes
PowerShell 7.7 includes a small number of breaking changes. Review them before you upgrade
production scripts.
New-Guid generates version 7 (time-sortable) UUIDs by default instead of version 4. Use
[guid]::NewGuid() when you need a fully random GUID.
Format-Table, Format-List, and Format-Custom now validate that -Property is not null or
empty.
Null-or-empty argument validation now throws System.ArgumentException.
Explicit false values for Where-Object operator switches, such as -EQ:$false, are now handled correctly.
Other behavior changes
These changes might also affect existing scripts:
The update notification feature was removed from PowerShell.
Test-Json requires -FetchExternalReference to resolve external schema references.
Export-Csv no longer accepts -Append together with -NoHeader. Using -NoTypeInformation
on the CSV cmdlets now writes a deprecation warning and has no effect.
Get-ChildItem -Name with -Recurse now returns matching files in subdirectories when the path
contains a wildcard or names a file.
Constrained Language Mode restricts property assignment on non-core objects, including through
ForEach().
Community contributions
PowerShell is built by a global community of users and contributors. Thank you to everyone who
filed issues, tested previews, improved docs, and submitted fixes during the PowerShell 7.7
release cycle.
We want to give a special thanks to the community members who spent the most time helping
others in the repository this release:
Top pull request commenters: @iSazonov, @muffy86, and @KirtiRamchandani
Top issue commenters: @mklement0, @237dmitry, and @rhubarb-geek-nz
The following individuals contributed code to the PowerShell 7.7 release:
We continue to work on future releases of PowerShell. Preview releases will continue to provide
early access to new capabilities and improvements, and your feedback on them helps us ship with
confidence.
Picture a service desk agent investigating a ticket. It pulls up the relevant policy, proposes a fix, and shows the technician exactly what it intends to change. The technician can edit the proposal, approve it, or send the agent back for another pass. That’s the .NET 11 demo I want to build. The interesting part is watching an agent and a person work on the same problem, with the person in control of what happens next.
Microsoft released .NET 11 on November 10, 2026, with runtime 11.0.0, SDK 11.0.100, and C# 15 as the default for projects targeting net11.0. The release brings faster developer tooling, better ways to model application states, and updates across ASP.NET Core, EF Core, MAUI, and Windows desktop. For teams working with Microsoft Foundry and Microsoft Agent Framework, the new Blazor AI components are the part I’m most excited to try. Microsoft’s .NET 11 launch announcement
Should you move from .NET 10?
If your .NET 10 application is humming along in production, you have time to make a good decision. Microsoft currently lists November 14, 2028 as the end-of-support date for both .NET 10 LTS and .NET 11 STS. The labels describe their support periods from release, and .NET 10 shipped a year earlier. Keep your .NET 10 deployments current on servicing and pick a reason to upgrade that you can explain to a customer. A faster startup, a better workflow, or a dependency you can finally remove makes a better case than a higher version number. Microsoft’s .NET support policy
If you’re on .NET 8 or .NET 9, the clock is different. Both reached end of support on November 10, 2026. Move to a supported release. .NET 11 is a good candidate when your hosts and dependencies are ready. Patched .NET 10 is a practical choice when it gets you back to a supported version sooner. MAUI has its own calendar. MAUI 10 support ends May 11, 2027, and MAUI 11 support ends May 9, 2028. If you ship mobile apps, put that first date on your migration calendar. Microsoft’s .NET support policy · .NET MAUI support policy
What developers get on Monday morning
C# 15 gives application states a better home. A union can represent the alternatives an operation returns, and a closed class hierarchy can let the compiler check whether a switch covers all permitted cases. Think of a workflow that can finish, fail, or pause for approval. When you introduce a new outcome, compiler checks can help you find the code that needs to handle it. C# 15 also adds collection expression arguments, extension indexers, and labeled break and continue. The new memory-safety rules are still in preview. What’s new in C# 15
The build and test loop gets faster. The SDK uses a Native AOT command-line host and a warm MSBuild server to reduce overhead. File-based apps can span multiple files with #:include, the CLI can manage solution filters, and dotnet watch gains crash recovery and Aspire integration. If your suite uses Microsoft.Testing.Platform, dotnet test adds run timeouts, failure limits, clearer progress, and per-module results. Try the new SDK against your current net10.0 target on a branch to see what your team gains from the tooling alone. SDK and tooling changes
The runtime and libraries can take work off your plate. JIT improvements help with operations such as bounds checks and virtual calls, and Native AOT improves interface dispatch. Built-in Zstandard compression, child-process output capture, in-memory stream adapters, and LINQ’s FullJoin give you more options for replacing custom plumbing. System.Text.Json adds support for C# unions and JSON Lines output from async sequences. The runtime libraries use Runtime Async internally, and enabling it in your own application code is still a preview feature. Benchmark the paths your customers actually use before putting a performance number in a release note. Runtime changes · Library changes
From a chat box to an agent people can work with
Back to that service desk. A chat transcript can tell a technician what the agent did. A live interface can show what it’s doing, what it proposes to change, and where it needs a human decision. The experimental Blazor AI components give you building blocks for that interface, including streaming chat, tool activity, approval prompts, and typed shared state. For example, the agent can update a proposed remediation plan while the technician edits a field in the same view. The application still owns authorization and decides when to perform the action. Microsoft says Microsoft.AspNetCore.Components.AI will remain experimental and prerelease throughout .NET 11. Microsoft’s Blazor agent UI walkthrough
Here’s how the pieces fit together in a Foundry-backed application:
Microsoft Foundry supplies the deployed model and can host your agent when you choose that hosting model. Microsoft’s guide to hosting Agent Framework agents in Foundry lists the .NET 10 SDK or later. Foundry hosted-agent guide
Microsoft.Extensions.AI gives .NET code interfaces such as IChatClient and IEmbeddingGenerator. They fit into familiar dependency injection and telemetry patterns and ship as NuGet packages. Microsoft.Extensions.AI overview
Microsoft Agent Framework coordinates the agent, its tools, and multistep workflows. Model Context Protocol (MCP) gives an agent a standard way to discover and call tools. The .NET 11 SDK includes a dotnet new mcpserver template using the official C# SDK, a quick way to try exposing an application capability as a tool. Microsoft Agent Framework overview · .NET 11 launch announcement
Agent User Interaction Protocol (AG-UI) carries events between a remote agent and the interface, including progress, tool activity, approval requests, and shared state. The standalone AG-UI .NET SDK is stable. The Agent Framework ASP.NET Core AG-UI hosting integration is prerelease at launch. AG-UI .NET SDK announcement · .NET 11 launch announcement
If you already run agents on .NET 10, try the new interface against one existing workflow with a real approval step. See whether a user can understand the proposed action, reject it cleanly, and reconnect without losing track of the work. For agents that run without a UI, keep an eye on the Foundry and Agent Framework releases themselves. For example, Agent Framework for .NET 1.28.0 removed workflow span attributes including executor.input, executor.output, and message.content. If your dashboards or evaluation pipeline read those fields, the package upgrade deserves its own regression run. Agent Framework .NET 1.28.0 release notes
What changes in web, data, and client apps
ASP.NET Core and Blazor. Async validation lets Minimal APIs and Blazor forms await a lookup. OpenAPI 3.2 is the default for generated documents and can describe server-sent event payloads. SignalR can refresh authentication on an existing connection. Blazor adds static rendering features such as CacheView, and Interactive Server apps can opt into pausing inactive circuits. Check custom OpenAPI transformers against the 3.x object model, test generated customer clients, and exercise form posts and reverse proxies under the new request-handling behavior. ASP.NET Core release notes · ASP.NET Core migration guide
EF Core 11. Query translation improves for joins, grouping, and supported full outer joins. Complex types and JSON columns work with more inheritance mappings. EF Core 11 requires .NET 11, so check your provider before pairing the upgrades. With Cosmos DB, saving an entity can drop properties absent from the EF model, and synchronous I/O is removed. Test against copies of real documents. SQL Server teams using affected Microsoft Entra authentication modes should check the separate SqlClient Azure extension dependency. EF Core 11 features · EF Core 11 breaking changes
MAUI and Windows desktop. MAUI moves to CoreCLR by default, adds preview incremental XAML Hot Reload, and changes the Windows CollectionView and Android Shell handlers. Check Android’s new minimum API level of 24 and the removal of 32-bit Android x86 support. WinForms offers opt-in modern rendering, and WPF gets fixes for window rendering, weak events, and self-contained publishing with mixed-mode dependencies. MAUI changes · WinForms changes · WPF changes · .NET 11 breaking changes
How I’d pilot the upgrade
Pick one app and measure it. Record build time, startup, memory, and the metrics your customers care about. For a web service, include p95 latency and error rates. For an AI app, time model calls, tool calls, and your own application separately so you can tell where any improvement came from.
Move the build and then the app. Install the SDK side by side, update global.json if you pin the SDK, and run your existing build. Then retarget to net11.0, preserve platform suffixes, and update compatible packages, workloads, and deployment images. Microsoft’s upgrade steps
Test what you’ll actually ship. Exercise the container, self-contained build, or framework-dependent deployment on representative customer infrastructure. The x86/x64 runtime now requires the x86-64-v2 instruction baseline, including SSE4.2 and POPCNT. Check older physical machines and VM CPU profiles early. Run regression tests against the real inputs your app handles, including database records, API payloads, and archives. Minimum hardware requirements · .NET 11 compatibility catalog
Give the first customers a reason to volunteer. Pick a customer group that will see a measurable gain or use a new feature. Agree on what success looks like, test rollback with any data changes in scope, and bring the results to the next group.
Bring your .NET team along
The Training Boss can help you migrate to .NET 11 and train the people who will build and run it. We can assess your applications and dependencies, plan a phased migration, and work with your team on the first deployment. Our hands-on training can cover C# 15, ASP.NET Core, Blazor, Microsoft Foundry, and Microsoft Agent Framework, with exercises grounded in the work your developers do. Bring an application you want to modernize or an agent workflow you want customers to use. We’ll help you figure out what to build first and how to get your team ready to ship it. Contact The Training Boss for a consultation
ReSharper 2026.2 brought Junie to Visual Studio as ReSharper’s first coding agent. In 2026.3, you can also pick Claude Agent or Codex in the same tool window. All three agents can use ReSharper’s refactoring skill to run refactorings such as Rename. This release also supports .NET 11 and C# 15, and the test runner now supports xUnit v3 4.0 and NUnit 5.0.
Many developers already use Claude Agent or Codex in a terminal or separate app. In ReSharper 2026.3, you can choose Claude Agent, Codex, or Junie in the AI Assistant tool window. Ask your chosen agent to write and edit code, run terminal commands, or explain unfamiliar parts of your solution – all without leaving Visual Studio.
Getting started
Install or update to ReSharper 2026.3, and make sure AI Assistant is selected in the installer. If you don’t see the AI Assistant tool window later, run the installer again and choose Install or Repair next to AI Assistant.
Open the AI Assistant tool window, sign in to your JetBrains Account if asked to do so, and choose Claude Agent or Codex from the agent list.
Send your first prompt. The first time you use Claude Agent or Codex, ReSharper asks for your consent before installing the selected agent, since it’s developed by a third party. If Node.js isn’t installed on your machine, ReSharper installs that too.
ReSharper connects these agents through the Agent Client Protocol (ACP), an open protocol for connecting coding agents to editors.
Junie, Claude Agent, and Codex can use ReSharper’s refactoring skill to rename symbols or move types to other namespaces instead of editing files one by one.
Junie Lite
Junie Lite, a free model tier for Junie, is now available in ReSharper to anyone signed in with a JetBrains Account, even without a ReSharper license.
Licensing and limits
What do I need to use Claude Agent and Codex in ReSharper?A JetBrains AI subscription. Claude Agent and Codex consume your JetBrains AI quota, as does Junie unless you use the free Junie Lite tier.
Can I use my own Claude or ChatGPT subscription, or my own API key? Not in this release. ReSharper doesn’t use personal Claude or ChatGPT subscriptions, and it doesn’t support bring-your-own-key (BYOK).
Is there a free AI trial? ReSharper no longer offers JetBrains AI Pro trials. If you’ve already started one, you can continue using it until it ends. Junie Lite is currently available as a free tier, with no trial required.
Regional restrictions on the availability of AI features in Mainland China apply.
Day-one support for .NET 11 and C# 15
Microsoft launched .NET 11 today at .NET Conf 2026, and ReSharper 2026.3 already supports it.
C# 15 language support
ReSharper supports C# 15, including union types, closed class hierarchies, collection expression arguments, extension indexers, and labeled break and continue statements. Code analysis, completion, navigation, and other features work with the new syntax. New inspections, quick-fixes, and context actions help you adopt the new syntax in existing code.
ReSharper also supports the memory safety redesign preview. It recognizes the new safe modifier and unsafe expressions, highlights missing or redundant unsafe contexts, and offers quick-fixes for them.
Readable runtime-async code in the decompiler
.NET 11 supports runtime-async methods, which the runtime handles directly instead of through compiler-generated state machines. Runtime-async is still a preview feature in .NET 11. When you decompile one of these methods, ReSharper now shows it as an ordinary async method, and the IL Viewer marks it with the async keyword. The same improvements are available in dotPeek 2026.3.
Code analysis and coding assistance
ReSharper suggests collection expressions in more places, including lists and immutable builders filled with foreach loops, AddRange, or individual Add calls. New inspections catch character values used by mistake as array lengths or indices, flag culture-sensitive operations and implicit string sorting, and find using statements that dispose of a task instead of its result. Another inspection suggests tuple syntax instead of verbose ValueTuple creation.
Parameter information now groups overloads that share the same leading parameters, so long overload lists are easier to scan while you write a call.
For C++, ReSharper 2026.3 adds a class layout view that shows how a class is arranged in memory. This release also brings format string checks for std::format, std::print, {fmt}, and Unreal Engine’s structured logging macros, new refactoring options, and new CUDA inspections. See What’s New in ReSharper C++ 2026.3 for the details.
Unit testing
The test runner works with the latest versions of popular test frameworks:
ReSharper now supports xUnit v3 4.0 and NUnit 5.0.
Test discovery is more reliable for MSTest 4 projects.
ReSharper applies Playwright settings (such as LaunchOptions) from your .runsettings file, so you can run browser tests with the browser window visible.
Microsoft.Testing.Platform support is enabled by default, so ReSharper finds tests for frameworks like TUnit without any setting changes.
Log in
