A'muse'ing

The biggest social media company is betting that you’re ready to let AI optimise every corner of your personal life. But is efficiency in our lives coming before the thought of privacy? Meta is putting that exact question to the test. 

Earlier last month, Meta, the parent company of Instagram, Facebook & WhatsApp, launched their flagship AI agent called Muse. Unlike the Chat-GPT-like chatbots that much of the world has become accustomed to,¹ Muse acts more like a PA: filling in forms, sending emails, paying for things and booking events, while constantly working in the background, even when you close the app. In order for Jolly, Muse’s cute bean-shaped mascot, to do all these admin-oriented tasks for you, you must grant it access to your personal digital sphere. This means handing over individual access to your emails, calendar, personal messages, documents, banking and finance records, health records and data and internet-enabled home appliances, with a string of more connections promised² in the near future. 

While some users have praised Muse for cutting down on unnecessary subscription bloat,³ identifying and cancelling unwanted recurring charges, others have been quick to point out its shortcomings. A Facebook Marketplace user found⁴ Muse had given multiple sellers his address without permission, agreed to a low-ball offer, and then falsely told a buyer he was home. Given the uptick in what feels like weekly news headlines of AI agents not doing what their big tech creator directed⁵ and going rogue, it strikes a difficult balance for those keen to make their lives as efficient as possible. 

Muse is currently available in the US, Canada and Mexico, but like most Meta products, will likely make its way into the Australian market. Currently, Australian law⁶ prohibits the secondary use of personal information, unless the individual has specifically consented to such secondary use. But, given how easy it is to accept the T&C’s and privacy policies of these huge social media companies, consent is a grey area. So how do you know when your data is being used for more than just buying Powderfinger tickets? 

Well, you don’t. By default, all Muse interactions can be used to train Meta’s other AI models, unless users manually opt out. And Meta doesn’t have the best track record at having users’ best interests in mind when it comes to data collection. A reminder of this couldn’t be more timely as, earlier this month, The Social Reckoning hit the cinemas, centred on whistleblower Frances Haugen, who in 2021 exposed Facebook (now Meta) as intentionally prioritising company profit over public safety through the design of its engagement-oriented algorithms. The film has been widely reported to have ‘flopped'⁷ at the box office in comparison to its academy award-winning predecessor, The Social Network. This will of course come as good news to Mark Zuckerberg as Muse reaches the home screen of phones across America, garnering over 9 million⁸ downloads since its launch on September 8. 

Meta has long been waiting to join the big leagues of AI, alongside tech rivals Google, OpenAI and Anthropic. Meta’s release of Muse and Muse Wonder, its new frontier in agentic models, finally gives it a foot to stand on in the AI race. But at a time when the current Australian government is looking to pull away⁹ from the hyper-personalisation and proven divisiveness of social media content,¹⁰ it seems that while, at least for now, Australians might be safe from tech giants accepting low-ball offers on their behalf, the safety of their data remains in doubt.¹¹ 

References: 

1. Bailyn, E. (2026, October 7). ChatGPT Usage Statistics: October 2026. First Page Sage. https://firstpagesage.com/seo-blog/chatgpt-usage-statistics/ 
2. PostFast. (2026, October 13). Meta Muse Connectors: The Full List, With Sources (Updated Weekly). https://postfa.st/blog/meta-muse-connectors-list 
3. Leuven, H. V. (2026, September 28). This li’l guy’s passion is fiscal responsibility. Morning Brew. https://www.morningbrew.com/stories/this-lil-guys-passion-is-fiscal-responsibility 
4. Nguyen, M. (2026, September 30). Meta’s Muse Sent a Stranger to a User’s Door. Its Permission Settings Explain Why. Memeburn. https://memeburn.com/metas-muse-sent-a-stranger-to-a-users-door-its-permission-settings-explain-why/ 
5. Nolan, B., & Shao, W. (2026, October 16). AI agents are going rogue, and the labs’ own training choices are pushing them to do so. Fortune. https://fortune.com/2026/10/16/ai-agents-are-going-rogue-the-labs-own-training-choices-are-to-blame-reinforcement-learning/ 
6. OAIC. (2023, March 10). Chapter 6: APP 6 Use or disclosure of personal information. OAIC. https://www.oaic.gov.au/privacy/australian-privacy-principles/australian-privacy-principles-guidelines/chapter-6-app-6-use-or-disclosure-of-personal-information 
7. Pulliam-Moore, C. (2026, October 12). The Social Reckoning is flopping at the box office. The Verge. https://www.theverge.com/entertainment/1009630/social-reckoning-box-office-other-mommy 
8. Danai Nesta Kupemba. (2026, October 19). Will developing nations be able to handle millions of AI agents? Rest of World. https://restofworld.org/2026/meta-muse-openai-dots-ai-agents-global-south/ 
9. Parker, C., Albury, K., & Weatherall, K. (2026, October 15). Australia’s ‘digital duty of care’ was just introduced to parliament. So what’s in the bill? The Conversation. https://doi.org/10.64628/AA.3qhfx47pn
10.  Wike, R., Smerkovich, M., & Fagan, M. (2026, October 1). Across the Globe, People Increasingly Say Social Media Is Harming Democracy. Pew Research Center. https://www.pewresearch.org/global/2026/10/01/across-the-globe-people-increasingly-say-social-media-is-harming-democracy/ 
11. Reuters. (2026, September 29). OpenAI apologises for Australian government website hack, pledges to rebuild trust. Reuters. https://www.reuters.com/legal/litigation/openai-says-ai-models-accessed-australian-government-systems-without-2026-09-29/