Meta brings Muse to Australia with local services still rolling out
Editorial Team
••📖6 min read
Meta’s Muse personal AI agent has launched in Australia, with some local shopping and booking connections to follow. Its browser and payment tools broaden its reach, while the company’s technical disclosures qualify its privacy promises.
Muse is available in Australia, with some local shopping, property and other integrations arriving after launch.
Browser and payment tools extend its reach beyond the announced partner roster.
Meta says users control permissions, but its technical disclosures distinguish credential protection from privacy against Meta itself.
Meta, the parent company of Facebook, launched its Muse personal AI agent in Australia on November 17, 2026, according to its announcement. The software can act through connected services to handle emails, schedules and reservations, but some of its advertised Australian shopping and booking experiences will become available later.
The launch extends Meta’s effort to put an assistant between users and everyday services, including purchases. That makes the scope of its permissions and data access consequential: users are delegating actions involving accounts and money, rather than asking only for answers.
An Australian expansion, with a staged partner rollout
Muse is available on iOS, Android, the web and through a Mac app, Meta says. Support for AI glasses is described as coming soon.
This is an expansion of an existing product. Meta began rolling Muse out in the United States in September, with Australia among its first international markets, according to a launch report citing Meta’s media release. That account also lists Google Contacts and commerce services Shopify, Stripe Link and Shop Pay among Muse’s connections.
Meta says Australian users can connect Gmail, Google Calendar, Instagram and OpenTable to manage emails, coordinate schedules, use saved content and reserve restaurant tables. Its local partners cover a wider range of proposed tasks:
Canva, a design and visual communication platform, for creating visuals and content. Meta says this connector is also available globally.
Domain and realestate.com.au for exploring property listings, and SWEAT for fitness goals.
Kogan for electronics and appliances, MECCA for beauty products, Officeworks for school and office supplies, and Ticketek for tickets.
ShopBack for comparing prices and earning cashback and rewards on flights, hotels and shopping.
The timing qualification applies to that roster: Meta says connector features will arrive progressively, with some experiences following launch. A separate account of Meta’s announcement makes the same distinction between access to Muse and availability of local services. These are company-described uses, not independently demonstrated completion of every transaction.
Microsoft is pitching Copilot Workspaces as a shared environment for people and AI agents. Team collaboration is already in early access, but personal Autopilot integration is still coming soon, with final agent pricing unannounced.
Editorial Team
Real-time voice conversations are also available on mobile, web and Mac. Spaces, a feature inside Muse, lets users describe a custom experience they want it to create, such as a school dashboard, wedding planner or trip planner.
The agent can work beyond named connectors
The partner list does not define Muse’s entire reach. In its September 8 technical account, Meta said it had been developing and using Muse internally since early 2026. It described a dedicated cloud computer for each user, with storage, computing resources and a browser, allowing the agent to run code, handle background work and coordinate subagents.
Meta says its built-in connectors use service providers’ APIs, the interfaces that let software communicate. Muse can also write custom connectors for services with suitable interfaces and use a browser to navigate websites and fill forms. Users can view its browser activity and take over.
For purchases, the September account described a partnership with Stripe Link and said Shop Pay was coming soon. It said payments for new merchants could use single-use card numbers restricted to a merchant, amount and validity period, with user approval. The later launch report lists Shop Pay among commerce connections but does not establish when that integration became usable in Australia.
Acting through a computer is an established competitive approach. OpenAI, the developer of ChatGPT, introduced ChatGPT agent on July 17, 2025, combining its earlier Operator web-interaction system and deep research tool. It described website navigation, calendar work and shopping tasks on a virtual computer, with permission required before consequential actions.
That announcement is a historical comparison; the launch page now labels itself outdated. It establishes precedent for the mechanism, rather than a current price or performance comparison with Muse. Neither company’s announcement supplies a like-for-like test of Australian shopping and booking tasks.
Meta’s company-reported Muse architecture separates the agent runtime from credential storage and Sentinel. Its disclosure says the design does not prevent Meta from accessing VM data when necessary to operate the service. Source: Muse.
Protecting credentials does not remove every privacy tradeoff
Meta says Muse Secure VM, its dedicated cloud environment, stores connected-service data and credentials while keeping passwords and payment methods out of the main agent’s view. Users choose access permissions, and sensitive actions such as purchases and sending emails require approval.
The technical account describes Sentinel, a separate permission system that the main agent cannot override. It evaluates connector actions and outgoing network requests against user policies. Approvals can be limited to an action, task or period; previously permitted or low-risk actions can proceed without repeated interruption. Meta also says its email connector filters out one-time passcodes and password-reset links.
Meta says it tested its safeguards internally and with outside security researchers, and opened a public bug bounty program in September. Those measures do not establish that compromise is impossible. Meta explicitly says Muse can make mistakes and that malicious instructions in material it reads—known as prompt injection—remain an open problem.
There is also a relevant client-app incident. In September, Patrick Wardle, a macOS security researcher and co-founder of the nonprofit Objective-See, demonstrated a flaw that could let a local process redirect Muse’s dictated audio and potentially abuse access granted to the app. An updated report carried Meta’s statement that it had issued a hotfix. Meta stressed that exploitation required malicious code already running under the user’s account; it was not a remote exploit.
The cloud design has separate privacy limits. Meta’s September technical disclosure said operational policies restricted staff access, but the architecture did not prevent Meta from accessing VM data when necessary to support, secure or operate the service. It described a future Muse Confidential VM intended to cryptographically prevent such access. The Australian launch announcement does not establish that this planned protection has shipped.
Meta also says conversations and VM data are not shared with its advertising systems. It cautions, however, that browsing and transactions performed on a user’s behalf can indirectly affect the ads they see. Separately, it says interaction records—including conversations, tool calls and subagent handoffs—are sanitized to remove key personally identifiable information before use in model training. Users can opt out in settings.
The next test is what Australians can actually delegate
Meta describes Muse as free for most needs, with subscriptions for heavier use. The Australian announcement gives no subscription prices, usage allowances or delivery date for each local connector. It also provides no measured Australian task-completion rate or evidence of savings from delegated shopping and bookings.
Those gaps prevent a meaningful cost comparison: the existence of a free tier does not establish the cost of repeated, resource-intensive tasks on a dedicated cloud computer. The next evidence needed is a service-by-service availability record, clear paid limits and results showing which tasks finish successfully, with how much user intervention.
For users deciding whether to connect an inbox, calendar or payment service, those results must be weighed alongside the access granted and the data policies already disclosed. For Meta, publishing clearer limits and performance evidence would make the case for delegating those tasks easier to assess.
Google is rolling out file references and sharing for Gemini skills as adult personal accounts lose the ability to use Gems. Automatic conversion preserves supported material, but file limits and missing tools can interrupt existing workflows.
Muddy Waters disclosed a short position in Nvidia while questioning a possible connection involving senior adviser Ajay K. Puri. Its own account leaves the key identity question unanswered, separating its new claims from an existing server-diversion prosecution.
Nvidia reported $115.8 billion in quarterly revenue in its first earnings release since co-founder Jensen Huang resigned. Strong demand gives interim CEO Colette Kress a financial cushion, while a campaign for Huang’s return collides with unresolved investigations and the board’s reported opposition.
OpenAI launched GPT-6.2 Astra and Sol for professional work, cutting Astra’s standard input and output token prices by 20%. The models offer cheaper paths through selected tasks, but research evaluations, reporting failures and unfinished independent safeguard testing complicate the case for broader autonomy.
Microsoft’s first live speech-to-speech model offers OpenAI API compatibility and planned enterprise integrations. Its published evaluations show a low-cost option, with several rivals completing more agentic tasks.
Canberra-based Trellis Data is offering its 220-billion-parameter Agentaus 1 model to business and government. Its pitch goes beyond keeping data in Australia, but performance, deployment costs and a planned capital raising remain key tests.
Perplexity is adding direct collaboration between frontier models to Computer, letting a lead agent coordinate up to 16 supporting agents. Company-reported benchmarks suggest mixed teams can outperform repeated copies of one model, but leave customer costs and production reliability unresolved.
Marsh reported 105 client cyber incidents in the last two weeks of October, against 48 in the final two weeks of May. The broker links the increase partly to AI, while warning that insurance panel access may not guarantee an available response team.
Microsoft and Salesforce are extending their sales software alliance with Copilotforce, combining CRM workflows with Microsoft 365 context under existing permissions. The Copilot plugin enters public preview, while Teams remains in pilot; earlier Microsoft tools and a Claude beta raise the bar for proving its value.
Microsoft expands MAI Image 2.7 from an editing preview into image generation and API access, with cheaper Flash pricing and a Canvas workspace. Benchmark claims depend on the workload, while the MAI Labs interface carries restrictions on commercial use.
OpenAI is widening applications for its legal AI configuration after improving research scores and expanding its search index. Access still requires approval, and benchmark results show substantial limits alongside the gains.
Microsoft has made Word, Excel and PowerPoint creation generally available inside Copilot, combining editable files with workplace permissions and collaboration. The broader shift toward autonomous work remains staggered across previews, while its claimed cost advantage rests on company-funded testing.
Knit Health’s NEA-led Series A funds deployment of Med-BEAM for care navigation and hospital operations. Its early company benchmark beats three Claude models at predicting clinical actions, while better patient outcomes remain unproven by that test.
Science Corporation’s Series D follows its European retinal-implant launch. Clinical results support improved central vision, but U.S. approval, eligible patients and commercial terms remain consequential questions for the startup, reportedly valued at $3.5 billion.
Google will let Gemini users save conversations while opting out of generative AI training on future chats. The global rollout also extends storage of history-off and temporary chats from 72 hours to around two months; human review and longer retention can still apply.
Google is removing the subscription requirement for Gemini’s task mode in six countries, while adding U.S. payment, calling and messaging features. The expansion broadens access to delegated work, but usage limits, permissions and website compatibility still govern what it can finish.
Eligible Pixel owners can join a Play Store beta to generate home-screen widgets from descriptions, but device, language and regional restrictions still apply.
Microsoft’s new in-house coding models compete with outside suppliers inside GitHub Copilot. Company tests show lower task costs at promotional prices, while its continuing financial ties to OpenAI complicate the rivalry.
Valkryn disclosed a $15 million seed round and early military testing contracts. Its evaluator-focused Battle Lab faces existing commercial tools, public testing capabilities and funding constraints that software alone cannot resolve.