Meta’s Muse AI Agent Has a Security Hole That’s Hard to Ignore
Synopsis
Meta Muse AI agent faces a reported macOS zero-day that can expose account authentication, while Amazon blocks shopping access amid wider scrutiny of AI-agent security.
Meta Muse AI agent is insecure, as security researcher Patrick Wardle showed that its authentication token could be extracted in a zero-day vulnerability, allowing control of a user's Muse account.
The bug affects the macOS app, letting locally run apps or terminal commands modify an undocumented setting to change the location of speech for transcription in Muse. The traffic can be diverted to an attacker's server, potentially revealing the token and giving them control of the Meta Muse AI agent.
Muse’s broad permissions add to the risk
The flaw in the Meta Muse AI agent is being made public and could be used to target pre-installed permissions granted to the assistant. Wardle demonstrated proof-of-concept attacks that write files to a Mac, and take photos through the camera without any prompts.
Earlier this September, Meta launched Muse, a new assistant that can help with email, calendar, forms, customer service and shopping. The macOS app will also offer integrations for WhatsApp and social media accounts.
The launch came as Amazon announced it would block access to the Meta Muse AI agent when users want to use it to shop online. Third-party applications that sell goods on behalf of customers should respect the preferences of the service providers and follow their guidelines on whether to participate in such sales.
AI agents remain an early security frontier
The AI agent for Meta Muse comes as regulators take a closer look at the security implications of software that can perform tasks on behalf of users.
Respondents agreed that AI agents introduce novel security challenges and that cybersecurity practices need to be updated, according to a May 2026 report by the National Institute of Standards and Technology.
In the UAE, generative AI adoption climbed to 64%, compared to 60.9% in Singapore and 28.3% in the United States in the second half of 2025, according to Stanford's 2026 AI Index Report.
Australia was at 36.9% and the United Kingdom was 38.9%. The figures reflect the use of generative AI, but not AI agents.
Meta's revenue grew 28% year over year to $60.80 billion in the second quarter of 2026. Capital expenditure was $31.08 billion, and capital expenditure is expected to be $130 billion to $145 billion for the full year 2026.
Source: ars Technica
Pooja Malik is a business journalist with over six years of experience covering startups, entrepreneurship, and emerging trends. She has previously worked with leading media platforms such as YourStory Media and BW BusinessWorld, where she reported on business, policy, and market developments. Currently, she serves as Editor at The Inspirepreneur Magazine, where she writes and edits stories across business, lifestyle, and travel, with a focus on clarity, accuracy, and reader relevance.
You Might Also Like
Fei-Fei Li’s AI Startup Bags $1 Billion for 3D Intelligence
Dropbox’s Strategic Evolution: From Freemium to Enterprise Powerhouse
