TLDR: Push Security documented fake event registration pages for OpenAI, SpaceX and Anthropic that tell visitors to sign in with their work account to claim a reward, such as three months of ChatGPT Pro.[1] “Continue with Microsoft” generates a device code and opens Microsoft’s real device login page. The employee pastes the code, signs in, approves, and Microsoft issues tokens to the attacker’s session.[1][2] “Continue with Google” runs through a reverse proxy that captures the password, the MFA response and the session cookie.[1] The domain check, the padlock and the MFA prompt all pass, because the sign-in is real. For MSSPs, the useful response is to block device code flow where clients do not need it, move high-value users to phishing-resistant MFA, and teach one sentence: never type a code you did not ask for into a sign-in page.
On November 3 at 8:26 AM, an email landed in my inbox from “SendGrid Login Services.” The subject was “Your OTP Code.” The body had six digits and a helpful line underneath: “Didn’t request this? Manage your account login history here.”
It came from a domain that has nothing to do with SendGrid. It was one of five SendGrid-branded fakes sent to the same address between November 2 and November 5. The spam filter caught one. Four reached the inbox. “API Endpoint Is Not Responding.” “Your batch API requests are failing.” “Reminder - Our API Is Changing.” The OTP one is the cleverest, because it uses a code to make you worry and a link to make you act.
I had not requested a code, and that is the whole analysis. A code you did not ask for is a reply to a conversation you never started, the same lesson my own product taught me with a password reset in October.
Two days later, researchers published a campaign that turns that lesson around. In this one, the attacker does not send you a code to make you click. The attacker hands you a code and asks you to type it into a page you can trust.
The invitation
In late October, Push Security detected and blocked visits to fake event registration pages on its customers’ browsers.[1] The pages advertised an OpenAI Academy livestream called “ChatGPT for Social & Digital Marketing Teams,” an invite-only SpaceX briefing on November 12, and an Anthropic event at AWS re:Invent. Each one offered a reward for registering: three months of ChatGPT Pro, three months of Grok Heavy, or a “SpaceX share.” Each one said a work account was required.[1]
The delivery chain leaned on names people trust. Invitation links went through Google short links or Zapier’s link shortener, and earlier versions of the campaign arrived from GoTo Webinar’s mail sender or from real Zoom event pages.[1] A Cloudflare challenge asking visitors to “Confirm you’re not a robot” sat in front of the lure, which stops most automated scanners from ever seeing it.[1][2]
Push says the campaign targets marketing staff, and the lure is well chosen. People want access to the newest AI tools, and “sign in with your work account to register” is a sentence they have seen on legitimate event pages many times.
To be fair to the attackers, the kit has a copy-paste problem. Push found that the OpenAI page still sends Google sign-ins through a domain named after a Claude webinar, and an earlier page carried a code comment reading “Same as Claude."[1] Even phishing operators are struggling to keep their AI vendors straight.
Continue with Microsoft
This is the part that makes the campaign worth an essay.
When the visitor clicks “Continue with Microsoft,” the page asks a phishing-as-a-service platform called Kali365 for a Microsoft device code and shows it in a “Confirm access” dialog with steps to copy it.[1] An “Open Microsoft” button opens Microsoft’s real device login page, on Microsoft’s real domain, in a pop-up. The visitor pastes the code, signs in, and approves.[1]
Device code sign-in is a legitimate OAuth 2.0 flow, the device authorization grant.[2] It exists so that devices without a keyboard or a full browser, like a conference room display or a command-line tool, can be signed in from a phone or laptop. The device shows a code, the person types it on the real sign-in page, and the device receives tokens.
In this campaign the “device” is the attacker’s session. Microsoft issues the tokens to Kali365 without the employee ever handing over a password.[1][2] BleepingComputer notes that those tokens can reach the user’s resources and connected single sign-on applications.[2] Then the page shows “Signed in as …, opening Microsoft 365” and redirects the employee to a second, proxied Microsoft sign-in page that captures the password and session cookie as well.[1]
Walk through the checks a careful employee was taught. Is the URL Microsoft’s? Yes. Is there a padlock? Yes. Did MFA prompt me? Yes, and I approved it, because I was signing in. Every check passes because the sign-in really happened. The only wrong thing in the whole flow is where the code came from.
Continue with Google
The Google path is more familiar. A pop-up titled “Portal” opens a reverse proxy that relays Google’s real sign-in flow, and captures the password, the MFA response and the session cookie on the way through.[1] This is adversary-in-the-middle phishing, and the lookalike domain is the one thing a user could catch.
Push describes both services, Kali365 and VoidProxy, as rented platforms designed to bypass or defeat MFA checks, and the operator as a customer of both.[1]
Why “check the link” does not cover it
Most awareness programs teach users to inspect the sender and the URL. That advice still matters, and it would have helped at the event page, which ran on a domain that was three days old when Push investigated it.[1] It does not help at the step that hands over the account, because that step happens on Microsoft’s own page.
That is the shift worth explaining to clients. Device code phishing does not need a fake login page. It needs a person to complete a real sign-in on someone else’s behalf. The defense has to be a rule about the code, not about the page.
What changes for an MSSP
1. Block device code flow where nobody uses it
Microsoft recommends that organizations “get as close as possible to a unilateral block on device code flow,” and documents a Conditional Access policy that does it: target all users and all resources, set the Authentication Flows condition to device code flow, and block access.[3] Start the policy in report-only mode to see who actually uses the flow, exclude break-glass accounts, and allow it only for documented cases such as legacy tooling that cannot be updated.[3]
For most of your clients’ users, device code sign-in is something they have never done on purpose. That makes it a good candidate for a block that nobody notices.
2. Put phishing-resistant MFA where the session matters
The Google path, and the second capture on the Microsoft path, steal session cookies through a proxy. Phishing-resistant authenticators such as FIDO2 security keys are bound to the real site, so a proxy cannot complete the sign-in for them.[2][4] Prioritize admins, finance, executives and anyone whose mailbox could be used to phish the rest of the company.
Be honest with clients about what each control covers. Blocking device code flow addresses the Kali365 path. Phishing-resistant MFA addresses the proxy path. One without the other leaves a door open.
3. Teach one sentence about codes
The training message fits on a sticky note: never type a code you did not ask for into a sign-in page. BleepingComputer’s advice to employees is the same in longer form: avoid entering Microsoft device authentication codes provided by unfamiliar websites, and be cautious of event invitations that require signing in with a corporate account to claim a reward.[2]
It pairs with the rule from my SendGrid morning. If you did not start the sign-in, the code is not yours. It does not matter whether the code arrived in an email or on a web page, or how real the page that asks for it looks.
4. Rehearse the lure before it arrives
Event invitations with a reward attached are an easy pretext to rehearse. The page looks like a real event, the reward is something people want, and the ask is a sign-in. Run it as a simulation and look at what happens after the click: who would have kept going to “sign in with your work account,” and who stopped and reported it.
5. Count the reports
When one employee reports a fake event invite, every other recipient of the same campaign benefits, but only if someone is looking at reports. Make the report a single action, and measure how fast reports come in next to how many people clicked.
What not to over-read
Push Security sells a browser security product and documented this campaign from blocks on its own customers’ browsers.[1] Neither Push nor BleepingComputer published a count of victims, and nothing here says how many organizations were affected. The operator rotates domains every few weeks, so specific indicators age quickly.[1] What lasts is the technique: a real sign-in, completed by the victim, for a session that belongs to someone else.
Where Symbol Fits
Symbol helps admins and MSSPs run security awareness across many tenants without staffing a person for every client.
Our phishing simulation library is built from active campaigns, and custom templates are available on request, so a fake event invitation is a pretext you can put in front of a client’s users in a controlled way. Security awareness training is where the “never type a code you did not ask for” rule lands, next to MFA and reporting. And Report a Phish gives every user one action to send a suspicious invitation to your analysts, with an Outlook button or a forwarding address for other mail clients.
None of those five SendGrid emails needed a link opened to be judged. The code alone gave them away. Make sure your clients’ users can say the same about the next code someone hands them.
References
- Luke Jennings, “Analyzing a fake event campaign combining AiTM and device code phishing payloads,” Push Security, November 5, 2026. https://pushsecurity.com/blog/ai-themed-fake-event-campaign-analysis
- Lawrence Abrams, “Fake OpenAI, SpaceX events used to steal Microsoft, Google accounts,” BleepingComputer, November 5, 2026. https://www.bleepingcomputer.com/news/security/fake-openai-spacex-events-used-to-steal-microsoft-google-accounts/
- Microsoft Learn, “Block authentication flows with Conditional Access policy,” last updated April 7, 2026. https://learn.microsoft.com/en-us/entra/identity/conditional-access/policy-block-authentication-flows
- Cybersecurity and Infrastructure Security Agency, “Implementing Phishing-Resistant MFA,” fact sheet. https://www.cisa.gov/sites/default/files/publications/fact-sheet-implementing-phishing-resistant-mfa-508c.pdf
