The article warns that small and mid-size businesses are facing a squeeze: new risks from AI agents connected to company systems, and faster, more effective versions of familiar scams like phishing. It highlights real-world trends such as QR-code phishing and “ClickFix,” where fake error pages trick users into pasting commands into their own devices. It also notes that attackers can hide malicious instructions in content AI agents retrieve (like web pages or emails), potentially causing data exposure without a user clicking a link.
Attackers are combining familiar social-engineering tactics with AI to make old scams faster and more convincing. Two patterns stand out. First, QR code phishing asks employees to scan a code embedded in an email, often framed as a need to verify or restore access after a supposed login issue, which leads to a fake sign-in page designed to capture credentials. Second, ClickFix scams present a fake error message that instructs a user to paste a command into their own terminal, a tactic now frequently dressed up as AI troubleshooting and hosted on public AI sharing pages to appear legitimate.
These scams work because they exploit trust in routine workflows: verifying access, fixing a technical error, or following AI-generated guidance. QR codes sidestep the usual scrutiny given to links, since the scan happens on a phone rather than in a browser where hover-over checks are common. ClickFix relies on urgency and technical framing to get users to act before questioning the source. AI is compounding the problem by making phishing emails significantly more effective, with automated phishing reportedly achieving a much higher click-through rate than standard attempts, letting attackers scale these techniques with less manual effort.
A less visible risk involves AI agents connected to business systems. Because these agents act through permissions granted by their owner, they are vulnerable to indirect prompt injection, where an attacker hides commands in a webpage, email, or other resource the agent is instructed to fetch. This can potentially expose data without any employee clicking a malicious link.
Mirage safely runs attacks like this one against your own team, so you find out what happens before a real adversary does.
ClickFix is a scam where a fake error message tricks a user into pasting a command into their own terminal, and it is now often disguised as AI troubleshooting hosted on public AI sharing pages.
QR code phishing bypasses normal link scrutiny because employees scan a code with a phone instead of clicking a link, often under pressure from urgent language claiming a login issue needs fixing.
AI-automated phishing emails reportedly achieve a 54 percent click-through rate compared to 12 percent for standard attempts, meaning attackers can scale proven social-engineering tactics more effectively.
AI agents can be manipulated through indirect prompt injection, where an attacker hides commands in a webpage, email, or other resource the agent is instructed to fetch, potentially exposing data without a user clicking anything.
AI scams are getting sneaky. The new trick? They make YOU do the dangerous part yourself. First, the email: 'Scan this QR code to verify your Microsoft 365 access.' You scan it, a login page pops up, looks normal, but the URL isn’t your real company sign-in. That’s QR code phishing, and it’s soaring. Then there’s ClickFix: a web page says, 'Error: Your session failed. Fix it by running this command in your terminal.' It’s dressed up as AI troubleshooting, hosted on a public AI sharing page, and the command can quietly install malware. Here’s the move: if an email wants you to scan a QR code, or any web page tells you to paste a command, stop and go through our official IT or portal instead, do not follow those on-screen instructions.
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.
The article describes real intrusions where attackers impersonate a company helpdesk and lure employees into "passkey, MFA, or SSO setup" steps. Victims are sent links via text (often to personal phones), leading to account takeover through adversary-in-the-middle phishing or device-code…
The bulletin describes real-world social engineering where staff were tricked into disclosing sensitive data or access. In one case, Revolut employees responded to fraudulent information requests sent from a real government-domain email account, exposing extensive customer records. Separately,…
Researchers observed a real phishing campaign that abused Microsoft 365’s Direct Send feature to make emails look like they came from the victim organization’s own domain, without compromising an employee account. The campaign was timed to mimic human sending patterns during U.S. Eastern business…
Anthropic reports multiple real-world threat groups used Claude to support cyber operations, including credential harvesting and data theft across many victims. The report includes specific, simulation-ready lures such as a fake ESET NOD32 login portal that sends stolen passwords to Telegram and a…
Anthropic says it disrupted a Russian state-linked campaign that used Claude to continuously rebuild malware when security tools detected it. The group (GTG-20006, linked to Midnight Blizzard/APT29) ran phishing and other human-targeted schemes, including device-code token theft against Microsoft…
