Many believe a text message code makes an account safe.
After being invited to a November 4 livestream on using ChatGPT for marketing, signing up gets three months of ChatGPT Pro for free.
The page looks like the real OpenAI Academy events site, but it is a fake. Security researchers say it is one of several lookalike pages built to steal Microsoft and Google logins.
Push Security, a firm that tracks browser-based attacks, spotted the pages in late October 2026. It blocked visits to fake registration pages for an OpenAI Academy livestream, a private SpaceX briefing on AI for marketing, and an Anthropic session at AWS re:Invent.
Push said it detected and blocked visits to the pages on customers’ browsers, so real people were being sent there. It has published its findings so other security teams can watch for the same kit.
The OpenAI version copies the design of the genuine events site and promises registrants three months of ChatGPT Pro. The livestream it advertises is aimed at social and digital marketing teams.
None of the three companies is behind these invitations. There is no livestream, no briefing and no free subscription waiting at the end of the link.
Invitations reach people through links that look ordinary.
Researchers saw Google short links and Zapier’s link shortener used to pass victims on to the fake pages, which helps the message look safe in an inbox.
Earlier versions of the same operation used fake Zoom event pages between May and October. Push tracks the newer kit separately, with pages first seen on October 22.
Another firm, Cofense, has described a similar campaign using fake invitations to a private Anthropic preview. The pattern is the same: a big AI name, an exclusive invite and a sign-in page at the end.
How the theft works
After you register, the fake page asks you to sign in with Microsoft or Google. That step is where the damage is done.
Researchers say the pages run on two phishing-as-a-service platforms called Kali365 and VoidProxy. Between them, the attacks use device code phishing and a reverse proxy, two techniques designed to get around multi-factor authentication.
The operators built the fake pages in two versions, one dressed as OpenAI and one as SpaceX, using the same underlying kit with a different skin.
That makes them cheap to produce and easy to swap for whatever brand is in the news.
Device code phishing abuses the sign-in method made for smart TVs and printers, where you type a short code on another screen.
The attacker starts a genuine login, hands you their code, and once you enter it on the real Microsoft page, they are given access to your account.
A reverse proxy works another way.
The fake site passes your details to the real Google or Microsoft page, so everything you see looks normal.
While that happens, the attacker copies your session cookie, the small file that tells a website you have already signed in. With it, they can walk into your account without needing your password or a fresh code.
How it could play out
To see how it works in practice, picture a freelance marketer who gets an email about a free AI workshop. The branding looks right and the offer is too good to pass up.
She clicks the link, lands on a convincing OpenAI Academy page and hits register. The page asks her to continue with her Google account, and the sign-in screen that follows looks exactly like the one she uses every day.
She enters her password and approves the prompt on her phone, as she always does. Behind the scenes, a copy of her signed-in session has just gone to a stranger.
She may see an error or a thank-you page, closes the tab and gets on with her day. Meanwhile, the attacker starts reading her email.
Many people believe a text message code or an authenticator prompt makes an account safe.
It does stop most attacks, but these ones are built to work around it.
Text message codes are the weakest of the common options because they can be intercepted or handed over to a convincing page. App-based prompts are better, but still fall to the methods above.
Once an attacker holds a valid session or access token, the account treats them as you.
The code you entered did its job, just for the wrong person.
Security professionals recommend phishing-resistant sign-in, such as passkeys or physical security keys. These are tied to the real website address, so they refuse to work on a lookalike page.
Passkeys are built into most recent phones and browsers. Both Google and Microsoft let you switch them on in your account security settings.
Who is in the firing line
Reporting on the campaign says the lures were aimed at employees, and the offers suit people who use AI at work.
Marketing staff, freelancers and anyone with a work Microsoft or Google login are obvious targets.
The pitch could work on anyone, though. A student hunting for free AI tools, a shop owner who runs everything from one Gmail account, or a relative forwarding a free offer in a family WhatsApp group could all be caught out.
A hijacked email account is a master key.
Attackers can use it to reset passwords for banking, shopping and social media accounts, and to message your contacts from an address they trust.
Employers have the most to lose. A single stolen login can open shared drives, client files and internal chat.
It takes only one person to say yes. From that login, attackers can look for ways into colleagues and clients too.
Businesses can cut the risk by moving staff to passkeys or security keys and by blocking sign-ins from unmanaged devices. Short training that shows staff a fake event page works better than a long policy document.
Anyone who runs a small business should treat their own email account the same way. Turn on passkeys, keep recovery details current and don’t use the same login for everything.
How to spot a fake invite
Check the link before you trust the logo. Real OpenAI, Google and Microsoft sign-in pages sit on those companies’ own domains, so be wary of any event page that sends you through a shortened link first.
Hover over a link on a computer, or press and hold it on a phone, to preview the address before you open it. A preview that shows a short link or an unfamiliar domain is reason enough to stop.
Signing in with Microsoft or Google to register is a convenience on some genuine sites. On an invitation you did not expect, treat it as a warning sign.
Look at the offer itself. Three free months of a paid product, a private briefing and a single livestream date all push you to act before you think.
Treat unexpected invitations the way you would an unexpected parcel. You didn’t ask for it, so confirm who sent it before you open it.
Never type a code into a Microsoft or Google page unless you started that sign-in yourself. If a site tells you to visit a login address and enter a code it has given you, close the tab.
When in doubt, go to the source. Open a new tab, type the company’s address yourself and look for the event or offer there.
What to do if you already clicked
Change your password from a device you trust, then sign out of every device and session you don’t recognise. Changing the password alone may not remove someone who is already signed in.
Run Google’s Security Checkup or review the recent activity page on your Microsoft account, both of which list devices and sign-ins. Look for anything you did not set up.
Attackers often add an email forwarding rule or a new recovery number so they can get back in later. Check both, and remove anything you don’t recognise.
If it was a work account, tell your IT team straight away. They can revoke access tokens and see what was opened, and a quick report limits the damage.
In the UK, you can forward suspicious emails to [email protected], the National Cyber Security Centre’s reporting service. If you have lost money, report it to Action Fraud on 0300 123 2040, or to Police Scotland on 101 if you are in Scotland.
Fraudsters follow attention, and AI tools have plenty of it. A fake invite carrying the ChatGPT, Claude or Gemini name gets clicks that a bland offer would not.
The tactic keeps changing its costume. In July, researchers at ANY.RUN reported lures using genuine Zoom event pages dressed up as partner summits for Meta, OpenAI and Anthropic.
The method underneath stays the same: a trusted brand, a reason to hurry and a sign-in page that is not what it seems. Spotting that pattern protects you, whatever name is on the invite.
Talk to family too.
Older relatives and teenagers who experiment with AI tools are both likely to see these offers, and a short chat now is cheaper than recovering a hacked account.
Free premium AI is exactly the sort of offer people want to believe. Spend thirty seconds checking where a link goes before you sign in.
If the invitation turns up in your inbox this week, don’t click through. Delete it and report it.
