Researchers found that a malicious instruction file placed in a user’s Downloads folder can manipulate AI coding agents into installing malicious software or exposing account login tokens. In testing, OpenAI Codex followed the planted instructions more than 90% of the time, while a related attack…
Researchers stopped a real attack in which a malicious instruction file in a developer’s Downloads folder attempted to direct an AI coding agent to install known malware. The attack abuses the agent’s trust in local instruction files and can also alter agent settings to steal session tokens or make…
Attackers used fake Microsoft sign-in pages and device-code tricks to steal completed Microsoft 365 login sessions, then replayed those sessions without needing passwords or new MFA approvals. After entering accounts, they registered rogue devices, replaced MFA methods, and hid mailbox messages,…
Attackers are abusing legitimate services such as Dell and Staples to send real, authenticated emails containing fraudulent support phone numbers. The emails pass normal authentication checks because the trusted platform sends them, but the attacker controls fields displayed in the message and uses…
This report describes real-world AI agent failures that exposed internal screenshots and personal bank information without a traditional attacker. It also documents a prompt-based technique that can bypass some AI safety controls and a repository-delivery flaw that can run code before a developer…
Proofpoint described real campaigns where attackers used highly tailored, AI-style messages to make phishing and payment fraud requests appear routine. Examples included property enquiries that led to malicious meeting invitations, device-code phishing pages disguised as business services, and fake…
Cofense analyzed real phishing emails disguised as bidding invitations. The emails send recipients to fake, branded login and device-enrollment pages, then trick them into entering a device code in a genuine Microsoft prompt. Doing so can authorize an attacker’s session without giving them the…
Researchers uncovered exposed infrastructure allegedly used by affiliates of INC Ransomware and The Gentlemen Ransomware. Alongside device exploits and ransomware tooling, the records describe a practical extortion workflow: direct emails to owners or executives, followed by voice calls, pressure…
An Arup employee transferred $25 million to fraudulent accounts after joining a video call featuring deepfake impersonations of company executives. The case shows how realistic-looking executive impersonation can pressure employees into approving high-value payments, even when the request appears…
Cybercriminals impersonated well-known technology brands including Claude, Google, Apple, and Docusign to steal credentials, payment details, or gain access to devices. The lures used urgent payment failures, software-update prompts, and promised device-performance improvements to persuade people…
A Hong Kong employee at engineering firm Arup transferred about $25.6 million after joining a video conference where every other participant was a deepfake. The incident shows how convincing fake video identities can be used to pressure employees into approving high-value payments, and why…
Researchers disclosed SearchLeak, a Microsoft 365 Copilot Enterprise vulnerability chain that reportedly let a single crafted-link click turn a user’s AI assistant into a tool for collecting and sending out sensitive data. The attack abused prompt injection and trusted Microsoft infrastructure,…
Attackers are using short business emails that ask companies to provide a quote for an upcoming project. The messages imitate ordinary request-for-quotation (RFQ) activity, then direct recipients to credential-stealing pages. The tactic is especially risky because sales and procurement teams…
A tracked threat group (UAT-11985) sent tailored event-invitation lures to people tied to Taiwan research organizations, trying to trick them into “registering” through a fake Google Form flow. Victims who clicked a disguised link or scanned a tampered QR code were pushed to a convincing Google…
Researchers observed a surge in short "request for quotation" (RFQ) emails that look like routine business inquiries to sales and procurement teams. After the victim replies, attackers send a follow-up link disguised as project documents or a procurement portal that leads to a fake Microsoft…
A weekly incident roundup describes multiple real-world attacks where criminals used "AI" branding to trick people into handing over passwords/MFA codes or running harmful PowerShell commands. The write-up includes two practical, repeatable social-engineering workflows: fake AI-platform login pages…
Security researchers showed that “decision models” like TypeSafe AI’s Jev can still be manipulated by indirect prompt injection, even if they don’t produce free-form text. In a shopping checkout test, a fake on-page ad persuaded an AI browser agent to click an attacker-controlled button instead of…
US authorities say North Korean operatives used stolen and fake identities to get hired into remote roles at more than 70 US companies, then operated company laptops from inside the US using “laptop farms.” Americans allegedly helped by receiving employer-issued laptops, keeping them online, and…
US authorities arrested two Americans accused of helping North Korean “fake IT workers” get hired remotely at US companies and route company laptops into US-based “laptop farms.” The scheme used stolen or fake identities and location-masking tools so overseas workers could appear to be working from…
North Korean-linked actors are using fake identities to get hired into remote IT roles, then leveraging that trusted access to steal data or attempt to deploy malware. The article describes real cases (including KnowBe4 and Netskope) and explains how the scheme scales using fabricated personas,…
Mirage simulates attacks like these against your own team, live and safely, so you can measure how your people actually respond.
