AI-Themed Phishing Lures in 2026: Fake Events, Fake Apps and How to Spot Them

Also on CyberFence: compare CyberFence to other VPNs side-by-side · the free CyberFence security tools .

Everyone wants the newest AI tool. Attackers know it. In the past few weeks, security researchers have documented two separate campaigns that use AI excitement as bait: fake invitations to exclusive AI company events that steal work accounts, and fake AI desktop apps that quietly install malware. Neither one needs you to fall for a clumsy email full of typos. Both look polished, both lean on trusted platforms, and both work because people are curious and in a hurry.

This post explains how each campaign works, why the usual "check the link" advice only goes part of the way, and the specific habits that protect you. We will also be clear about where a VPN helps and where it does not.

Trend 1: Fake AI Event Invitations That Steal Work Accounts

In late October 2026, browser security company Push Security detected and blocked visits to fake event registration pages that impersonated well-known AI and technology companies. BleepingComputer reported that the pages advertised livestreams and private product briefings and promised rewards such as three months of a premium AI subscription or a "share" in a private company.

The catch: every page said you had to sign in with your work account to register and claim the reward. Visitors saw two buttons, "Continue with Microsoft" and "Continue with Google." Both led to account theft, by two different routes.

The Microsoft route: device code phishing

Clicking the Microsoft button showed a short code and told the visitor to paste it into Microsoft's real device login page. That page is genuine. The sign-in happens on a real Microsoft domain, and the victim never types a password into a fake site at that step. But the code belongs to the attacker's session, so approving it hands the attacker a valid login token for the victim's account.

According to Push Security, the page then sent the victim on to a second, fake Microsoft sign-in page that captured the password and session cookie as well, giving the attacker two ways in.

The Google route: a reverse proxy

Clicking the Google button opened a pop-up that relayed Google's real sign-in flow through an attacker-controlled server. This is called an adversary-in-the-middle attack. The victim sees a normal login and a normal multi-factor prompt, and everything works as expected. Meanwhile the proxy keeps a copy of the password, the MFA response and the session cookie that proves the victim is logged in.

We covered why stolen session cookies are so dangerous in our guide to session hijacking and token theft. In short: once an attacker holds your session, your password and your MFA code no longer matter.

Why these pages are hard to spot

• Trusted link shorteners. Push Security found that invitation links passed through Google and Zapier short-link services before reaching the fake page.
• A real-looking bot check. A "Confirm you're not a robot" challenge sat in front of each page. It made the page feel legitimate and also blocked many automated security scanners from seeing what was behind it.
• Legitimate senders. Earlier versions of the campaign arrived through real webinar email systems and real event pages, according to Push Security. Cofense separately reported fake AI preview invitations sent through a webinar platform's mail service.
• Automatic translation. The pages translated themselves into the visitor's language.

This is not the same as a fake CAPTCHA that tells you to paste commands into your computer, which we explained in our ClickFix fake CAPTCHA guide, but it borrows the same idea: a familiar security step makes people relax.

Trend 2: Fake AI Desktop Apps That Install Malware

The second campaign targets people searching for AI software to download. Check Point Research tracked a campaign it calls Yoba for about two months. The attackers built GitHub projects and simple GitHub Pages sites that advertised desktop versions of brand-new AI models, using names that sounded like real upcoming releases.

Check Point says these pages likely ranked near the top of search results through search engine poisoning. Visitors downloaded a ZIP file. Inside was a genuine, digitally signed installer from a real AI company, plus one extra file: a malicious library that the real installer loaded automatically. The security term is DLL sideloading. Windows trusted the signed program, and the signed program loaded the malware.

Check Point is clear that the AI companies themselves were not hacked. The attackers simply repackaged real software. In the cases Check Point analyzed, the final payload was a cryptocurrency miner that paused while the user was active to avoid notice. But the loader is generic, and Check Point warns that the operator can swap in any payload, including password stealers or remote access tools. An earlier version of the same malware posed as a popular AI coding tool in June.

Why "it's signed" is not enough

Many people have learned to trust a valid digital signature. Normally that is good advice. Here, the signed file really was legitimate. The problem was what sat next to it in the same folder. A familiar product name, a GitHub download page and a valid signature can all appear in the same malicious package.

If a fake app delivers a password stealer instead of a miner, the damage goes far beyond a slow laptop. We explained what infostealers take and what a VPN can and cannot do about them in Can a VPN Stop Infostealer Malware?

Why AI Makes Such Effective Bait

Phishing has always followed whatever people are excited about: tax refunds in spring, package deliveries in December, and now AI. A few things make AI lures especially strong in 2026:

• Constant new releases. New AI models and features ship so often that a "new desktop app" or "private preview" sounds plausible. Hardly anyone can say for sure which releases are real.
• Work pressure. Many employees feel they need to keep up with AI tools for their jobs, so a work-account sign-in for an AI event does not seem strange.
• Free premium access. Paid AI subscriptions are valuable, so "three free months" is a reward people act on quickly.
• Better-looking fakes. Push Security notes that AI tools also make it easier for attackers to build convincing copies of real brand pages.

For a broader look at how criminals use AI to write the attacks themselves, see our post on AI phishing in 2026.

How to Protect Yourself From AI-Themed Lures

For event invitations

• Never enter a device code that a website or invitation gave you. A real device code flow starts on a device you own, such as a smart TV or a new laptop, and you know why you started it. If a web page hands you a code and sends you to a login page, stop.
• Treat "sign in with your work account to claim a reward" as a red flag. Legitimate event registrations rarely require a full corporate sign-in to hand out free subscriptions.
• Verify events on the company's own website. Type the company's address yourself and look for the event there. Do not rely on the link in the invitation.
• Do not trust a page because it shows a bot check or uses a short link. Both were part of this campaign.
• Use phishing-resistant sign-in where you can. Hardware security keys and passkeys are much harder to relay through a proxy than text codes or app prompts.

For AI app downloads

• Download AI apps only from the developer's own website or your device's official app store. A GitHub repository with a model name in the title is not an official source, no matter how high it ranks.
• Be suspicious of "leaked" or "early access" versions. If the company has not announced a desktop app, a site offering one is a warning sign.
• Think twice before granting administrator rights. According to reporting on the Yoba campaign, the malware asked for administrator access. Ask yourself why a chat app needs full control of your computer.
• Keep your security software and operating system updated so known malware families are caught faster.

If you already clicked

• Change your password from a device you trust, then sign out of all sessions so any stolen session cookie stops working.
• Check your account for new sign-in methods, forwarding rules or connected apps you did not add.
• If it was a work account, tell your IT or security team right away. Reporting fast limits the damage.
• If you ran a downloaded installer, disconnect from the internet and run a full malware scan before signing in to anything else.

Where CyberFence Fits, and Where It Does Not

We think honesty matters more than a sales pitch, so here is the plain version.

What helps: CyberFence Web Shield blocks known malicious and phishing domains at the DNS level, so once a lure domain is identified and added to blocklists, your devices stop connecting to it. That covers phones, tablets and computers on your plan, not just your browser. The CyberFence VPN also encrypts your traffic with AES-256-GCM, which protects your sign-ins and session data from snooping on public or shared Wi-Fi. CyberFence is US-operated and runs under a zero-logs policy.

What no VPN can do: A VPN cannot stop you from typing a device code into a real Microsoft page, and DNS filtering cannot block a malicious file hosted on a legitimate site like GitHub without blocking the whole site. Brand-new phishing domains may also work for a short time before any blocklist catches them. That is why the habits above matter. We go deeper on these limits in What a VPN Does Not Protect You From.

The best protection is layered: careful habits, phishing-resistant sign-in, up-to-date software, and network-level protection that blocks known threats before they load.

The Bottom Line

AI-themed lures are a trend, not a one-off. The fake event campaign ran with one toolkit from May to October 2026 and with a newer one from late October, according to Push Security, and the fake app campaign has changed its disguise as new AI products appear. Expect more of both as new AI releases keep coming. The rule of thumb is simple: if an AI offer asks for your work sign-in, a code you did not request, or a download from somewhere other than the company's own site, treat it as phishing until you have verified it yourself.

Want to go deeper? Read the CyberFence competitor comparison hub , our free privacy and security tools , or CyberFence plans and pricing .