Attackers are buying sponsored Google search ads that send Mac developers to fake OpenAI Codex download pages. The pages instruct victims to paste a Terminal command that looks like a normal install step, but actually downloads and runs a multi-stage malware payload.
Mirage safely runs attacks like this one against your own team, so you find out what happens before a real adversary does.
You Google “OpenAI Codex download for macOS” and click the top result, looks legit, OpenAI logo and all. But that click lands you on a Google Sites page impersonating the Codex download. Instead of a normal installer, it says: “Install Codex for macOS: open Terminal and paste the following command to complete installation.” The command looks like a normal npm install, but tacked on is obfuscated Base64, curl, and a pipe into zsh, quietly pulling a Mach-O payload into /tmp/helper and stripping macOS security flags. One paste, multi-stage infection. If any website tells you to paste a long install command into Terminal, stop. Instead, go directly to openai.com yourself and only follow install steps from there.
Attackers used a sponsored search ad to send macOS users to a fake “OpenAI Codex download” page hosted on Google Sites. The page convinced victims to open Terminal and paste a command that secretly downloaded and ran a multi-stage malware infection.
A real-world social engineering attempt used a legitimate Google Doc to trick a target into manually running commands that installed malware. The attacker posed as a crypto marketing executive and used a fake “decryption failure” message and a “manual update” button as the lure, leading to an…
The article describes how real-world phishing and social engineering are evolving to bypass the checks employees are trained to use (bad grammar, suspicious URLs, obvious fake login pages). It highlights specific, observed attack workflows including QR-code “device hop” phishing, OAuth token theft…
Attackers are sending fake ChatGPT billing emails that pressure people to “update payment” within 48 hours to avoid service interruption. The message links to a convincing look‑alike ChatGPT login page via a Google redirect, aiming to steal OpenAI credentials.
Unit 42 showed that default AWS AgentCore Harness settings can let an attacker use prompt injection to trick an AI agent into running shell commands and exposing plaintext credentials from AgentCore Identity at runtime. In their demo, a malicious support ticket embedded instructions (via hidden…
Researchers described AmnesiaStealer, a macOS info-stealer spread through a counterfeit “Download for macOS” page that tricks users into pasting a command into Terminal. The malware steals passwords and browser session data, and can even give an attacker live, hidden control of the victim’s browser…
