-
17:50
-
17:30
-
17:20
-
17:00
-
16:50
-
16:30
-
16:20
-
16:00
-
15:50
-
15:50
-
15:30
-
15:20
-
15:00
-
14:50
-
14:30
-
14:20
-
14:00
-
13:30
-
13:20
-
13:00
-
12:50
-
12:20
-
12:00
-
11:50
-
11:30
-
11:20
-
11:00
-
10:50
-
10:30
-
10:20
-
10:00
-
09:50
-
09:30
-
09:20
-
09:00
-
08:50
-
08:30
-
08:20
-
08:00
-
07:50
-
07:30
-
07:21
-
07:00
Chatgpt share feature abused to spread atomic macos stealer
Cybercriminals are weaponizing ChatGPT’s chat sharing feature in a new campaign that installs Atomic macOS Stealer, also known as AMOS, on Apple computers while posing as a guide to a fake “Atlas browser” for macOS. The operation targets a broad audience of macOS users and cybersecurity conscious professionals who rely on search engines and official looking resources to discover new tools and troubleshoot technical issues.
How the chatgpt share feature is abused
Security analysts report that attackers are publishing polished installation guides for a non existent Atlas browser as shared conversations hosted on the official chatgpt.com domain. These shared chats are stripped of suspicious context and presented as legitimate step by step instructions, making them appear trustworthy to users who reach them via search.
To drive traffic, the threat actors buy Google ads targeting terms such as “chatgpt atlas,” sending users to chatgpt.com/share URLs that look indistinguishable from genuine OpenAI content. Once on the page, victims are told to copy a single command into the macOS Terminal, framed as a standard installation step for the supposed browser.
From one line command to full system compromise
The command shown in the shared chat retrieves and runs a script hosted on the domain atlas-extension.com, which acts as the delivery mechanism for AMOS. The script repeatedly prompts the user for their macOS password and, once the correct credentials are provided, uses them to install the malware with elevated privileges and to set up persistence.
Researchers describe this social engineering method as a variation of the ClickFix technique, in which users are convinced to execute a command presented as a fix, update, or optimization rather than as a threat. In this case, curiosity around an apparently new ChatGPT linked browser is used to override normal caution, especially among users accustomed to pasting commands from technical guides.
Amos infostealer and its new persistent backdoor
Once installed, Atomic macOS Stealer focuses on harvesting sensitive data, including passwords, cookies, and autofill information from browsers such as Chrome and Firefox, as well as credentials and assets from cryptocurrency wallets including Electrum, Coinomi, and Exodus. The malware also searches Desktop, Documents, and Downloads folders for text, PDF, and DOCX files, and can capture session data from applications like Telegram Desktop and OpenVPN Connect.
Recent analysis shows newer AMOS variants ship with an integrated backdoor that gives attackers persistent, remote access to infected Macs even after reboots. Cybersecurity firm Moonlock notes this marks only the second globally scaled macOS backdoor operation after previous campaigns linked to North Korean groups, and warns that AMOS has already been detected in attacks across more than 120 countries, with the United States, France, Italy, the United Kingdom, and Canada among the most affected.
Global campaigns and blocked attacks
Threat intelligence reports describe AMOS as one of the most widespread macOS stealers currently in circulation, offered under a malware as a service model to different criminal crews. These groups combine techniques such as malvertising, fake support pages, and now shared AI chats to lure victims into running single line installation commands that bypass normal user skepticism and native security prompts.
CrowdStrike recently disclosed that its Falcon platform blocked more than 300 attempted intrusions between June and August 2025 involving SHAMOS, a variant of Atomic macOS Stealer developed by the cybercrime group Cookie Spider. Those campaigns relied heavily on fraudulent macOS help sites promoted via online advertising, illustrating how commercial off the shelf stealers like AMOS are being repeatedly repurposed and redistributed through evolving social engineering vectors.
Mitigation advice for macos users
Security specialists stress that users should never execute Terminal commands copied from unverified chats, search results, or online guides, regardless of whether the content is hosted on a trusted domain. Experts recommend checking any unfamiliar one line command with IT teams or by using independent analysis tools, and verifying software downloads through official vendor websites rather than search ads or intermediaries.
Defenders advise macOS users, including professionals managing sensitive data or cryptocurrency assets, to run reputable endpoint protection capable of detecting AMOS and similar stealers, to keep systems and browsers updated, and to monitor for unusual access to accounts and wallets. Organizations are also encouraged to train staff about ClickFix style tactics that disguise malicious commands as harmless fixes or productivity enhancements.