Fake Conference Invites Are Getting More Polished. Three Layers Keep Teams Ahead.

Key takeaways

• Cisco Talos tracked a fake conference-invite campaign by UAT-11985 targeting people tied to Taiwan research organizations, using respected academic institute names, Google Form look-alike pages, and in some cases swapped QR codes on genuine event posters.
• The attack used an adversary-in-the-middle login relay: The Hacker News reported the phishing page passed each credential entry and even the verification step to Google in real time, making the sign-in appear normal from start to finish.
• The article’s first defense layer is stronger sign-in: a 2025 UC San Diego study of 19,500+ health system employees across 10 phishing campaigns recommended technical countermeasures, especially two-factor authentication and password managers that only work on correct domains; passkeys and hardware security keys apply the same domain-checking principle.
• The second layer is AI-first inbox protection: Adaptive’s API-based email security for Google Workspace and Microsoft 365 uses behavioral signals, intent analysis, and LLM reasoning, returns a confidence score plus written explanation, and can remove confirmed threats from inboxes in seconds; the article cites Harvard Business Review authors Fredrik Heiding, Bruce Schneier, and Arun Vishwanath supporting LLM-based phishing detection when used correctly.
• The third layer is practiced user habits: type the organizer’s address yourself, verify where a QR code points before signing in, and report suspicious messages to a named channel; the UC San Diego study found click rates ranged from 1.82% for a fake Outlook password update to 30.8% for a fake vacation policy update, showing relevance strongly affects behavior.
• If credentials are entered, the article recommends a fast, blame-free response: reset the password, revoke sessions and tokens, review account activity including new mail rules and OAuth app grants, and thank the reporter; it cites 2024 University of Luxembourg SUPS research and Stanford’s Jeff Hancock on reducing shame to improve phishing reporting.

Want more from Adaptive in your Google results?

A Cisco Talos report on fake conference invitations shows why strong sign-in, AI-first email protection and everyday habits work best together.

Conference invitations are one of the best parts of working in security. They bring new ideas, new colleagues and a reason to step away from the inbox for a day. Saying yes is a good instinct.

Attackers know that, and Cisco Talos researchers recently documented a campaign built around it. The group, tracked as UAT-11985, emailed people tied to Taiwan research organizations with invitations to policy and geopolitical events. The messages carried the names of respected academic institutes. The links led to a page that looked like a Google Form and asked visitors to sign in with their Google account.

This sequence has three steps: an invitation that earns a click, a familiar-looking page, and a sign-in that appears to work normally. Each step has a practical answer. Teams that combine stronger sign-in, AI-first email protection and well-practiced habits can keep accepting the invitations worth accepting.

What the researchers found

Talos describes a message in three parts. It opens with context on a policy topic, adds a personal compliment and closes with event logistics. Researchers believe AI helped produce the emails at scale, since their structure matches so closely. In some cases, the group also swapped the QR code on a genuine event poster for its own.

The link led to a Google-style sign-in page that worked as a relay. According to The Hacker News, it passed each entry to Google as it was typed, including the verification step, so the person saw a normal login from start to finish. Security teams call this an adversary-in-the-middle page: a look-alike that sits between a person and the genuine sign-in and repeats every step.

Talos saw the activity aimed at one community of researchers. Polished invitations, QR codes and live relays can reach any industry, so the lessons apply to every team that signs in to cloud tools.

Why the invitation works

Familiar names. Familiar tools. Familiar rhythm. Sender names match institutions people already respect, topics match the work recipients do, and the compliment reads like a note from a thoughtful organizer. Even a careful hover over the link shows a Google Forms address.

Carnegie Mellon’s Lorrie Cranor, who directs the university’s CyLab Security & Privacy Institute, argues that security systems should spare people from tasks “they shouldn't have to be doing in the first place.” Spotting a flawless forgery by eye fits that description. The strongest programs build protection around people, so no single click carries the full weight.

Layer one: sign-in that checks the address

Start with how people sign in. A 2025 study from UC San Diego followed more than 19,500 health system employees across 10 phishing campaigns. Its authors recommend that organizations “refocus their efforts to combat phishing on technical countermeasures,” and they name two with the best return: “two-factor authentication for hardware and applications, as well as password managers that only work on correct domains.”

Passkeys and hardware security keys follow the same principle. They check a website’s address before they sign anyone in, so a look-alike page cannot complete the login. Admins can begin with executives, finance teams and anyone who works with outside organizations, then widen the circle.

Policy settings round out this layer. Admins can turn off one-time-code fallbacks for high-visibility roles, apply conditional access rules that check device and location, alert on unusual sign-ins and review new OAuth app grants. Each setting narrows the paths open to an attacker after a password is typed.

Layer two: AI-first protection at the inbox

Better detection shrinks the number of messages that ever reach a person. Adaptive’s agentic email security connects to Google Workspace or Microsoft 365 through an API and reviews each inbound message using behavioral signals, intent analysis and LLM reasoning. Every verdict carries a confidence score and a written explanation. Confirmed threats come out of affected inboxes in seconds.The approach has academic support. Harvard researcher Fredrik Heiding and Harvard Kennedy School lecturer Bruce Schneier, with co-author Arun Vishwanath, wrote in the Harvard Business Review that LLMs “can be helpful for detecting and preventing phishing emails, given they are used correctly.” Their related study found the models showed “a strong ability to detect malicious intent, even in non-obvious phishing emails.” Intent stays constant as lures grow more polished, which is why reading each message in context matters.

Setup takes minutes and needs no mail-routing changes. Email protection covers what reaches the inbox. A poster with a swapped QR code, or a link that starts on a legitimate service, benefits from the sign-in and habit layers working alongside it. The three layers cover each other’s gaps.

Layer three: habits, practiced together

Three habits take seconds each. Type the event organizer’s address yourself. Check where a QR code points before signing in. Report anything curious to a named channel, where the security team replies with a quick thank-you.The UC San Diego study adds a useful data point on practice. Across 10 campaigns, click rates ranged from 1.82% for a fake Outlook password update to 30.8% for a fake vacation policy update. That gap shows relevance changes behavior more than anything else in the study. It supports the idea that practice should match the messages people see every day.

Adaptive simulations mirror the lures attackers use today across email, text and voice calls, and each person gets feedback that fits their role. Employees get to try each habit once before they need it.

If a password has already been entered

Teams recover fastest when the response is quick and blame-free. Reset the password, revoke active sessions and tokens, and review recent account activity, including new mail rules and app grants. Encourage and thank the person for speaking up. Early reports shorten every response.Research supports that approach. University of Luxembourg researchers, presenting at the 2024 Symposium on Usable Privacy and Security, found that a lack of feedback and communication, among other factors, discourages employees from reporting phishing emails. A quick thank-you takes seconds and addresses that gap. Stanford professor Jeff Hancock described the goal as helping businesses “de-shame the reporting of mistakes.”

Three checks for this week

Security leaders can confirm the foundation in three steps.

1. 1Executives, finance and IT admins sign in with passkeys or hardware keys, and fallback methods are turned off for them.
2. 2Alerts exist for unusual sign-ins and new OAuth app grants, and a named person owns the response.
3. 3A one-page playbook covers a password entered on a suspicious page, and every employee knows where to report.

Saying yes with confidence

Great invitations will keep arriving, and teams can keep accepting them. Strong sign-in removes the main path a look-alike page relies on. AI-first email protection trims what reaches the inbox. Practiced habits cover everything in between. Each layer carries part of the load.

Adaptive partners with security teams on all three. To see agentic email security and personalized simulations in action, visit adaptivesecurity.com.