A device code phishing campaign is using fake OpenAI, SpaceX and Anthropic event pages to take over Microsoft and Google work accounts. Push Security reported it on November 5, 2026. The pages offer AI rewards, such as three months of ChatGPT Pro, to anyone who registers with a work account.
The Microsoft version of this attack is hard to spot because the victim types a code into Microsoft’s real sign-in page. The Google version uses a proxy that relays Google’s real sign-in. Both give the attacker a working session, even when the account has multi-factor authentication. This post explains how each path works, then how Microsoft 365 and Google Workspace admins can shut them down, and what to tell staff.
The short version: in Microsoft Entra, block device code flow with a Conditional Access policy, starting in report-only mode. In Google Workspace, move admins and other key staff to passkeys or security keys. Tell staff never to enter a sign-in code they did not start themselves. If someone already signed in through one of these pages, revoke their sessions right away.
What Push Security found
Push Security’s analysis, written by Luke Jennings, says Push detected and blocked visits to these pages in late October 2026. It describes three lures:
- OpenAI: an "OpenAI Academy" livestream for marketing teams on November 4, with three months of ChatGPT Pro as the reward.
- SpaceX: an invite-only "Starship Product Briefing" on November 12. Visitors could pick three months of Grok Heavy or a "SpaceX share."
- Anthropic: an "Anthropic at AWS re:Invent" page, which was no longer live when Push looked.
Every page asked the visitor to sign in with a work account. According to Push, invitations arrived through Google short links and Zapier links. Earlier waves used a webinar platform’s mail sender and real Zoom event pages. A Cloudflare "Confirm you’re not a robot" check hid the pages from most scanners.
Push says the operator changes domains every few weeks, so blocking a list of domains will not stop this device code phishing campaign for long.
How device code phishing works
Device code flow is a real Microsoft sign-in method. Microsoft’s authentication flows documentation says it "lets you sign into devices that lack local input devices, like shared devices or digital signage." A TV or a conference room device shows a code. You type that code on your phone or laptop, sign in, and the device gets access.
Device code phishing flips that around. The attacker starts the sign-in, gets the code, and hands it to you. When you enter it and sign in, Microsoft gives the access to the attacker’s session.
Push describes the Microsoft path in this campaign step by step:
- The visitor clicks "Continue with Microsoft." The page requests a device code from a phishing kit called Kali365 and shows it in a "Confirm access" box.
- "Open Microsoft" opens Microsoft’s genuine device login page in a pop-up.
- The visitor pastes the code, signs in and approves. Microsoft then issues tokens to the attacker’s Kali365 session.
- The page then sends the visitor to a Microsoft sign-in proxy run on VoidProxy, a second kit, to capture the password and session cookie too.
Microsoft is direct about the risk. Its documentation calls device code flow "a high-risk authentication method that can be part of a phishing attack." Its page on Microsoft-managed policies says device code flow "is rarely used by customers, but is frequently used by attackers."
Why MFA and passkeys do not stop the Microsoft path
In device code phishing, the sign-in itself is real. The victim is on Microsoft’s own page and completes their usual MFA prompt, or a passkey. What goes wrong is that the attacker’s session receives the result.
That is why the fix for device code phishing on the Microsoft side is to block the flow itself. Our post on passkeys explains why they beat codes and push prompts on fake sites. That protection works when the fake page sits on the wrong domain. Here the domain is correct.
How the Google path works
Push says the Google path in this campaign is not device code phishing. "Continue with Google" opens a pop-up titled "Portal" on an attacker domain. It runs VoidProxy, a reverse proxy that relays Google’s real sign-in pages. The victim sees real Google screens, enters a password and answers the MFA prompt. The proxy keeps the session cookie.
This kind of attack, often called adversary-in-the-middle, depends on the victim signing in through the wrong domain. Passkeys and security keys are tied to the real domain, so they do not work through the proxy. That makes them the right tool here.
Stolen session cookies are the same prize we described in our post on infostealers and session hijacking. Once an attacker has the cookie, the password and MFA have already done their job.
Block device code phishing in Microsoft Entra
Microsoft gives two routes. Both need Conditional Access, which Microsoft says requires a Microsoft Entra ID P1 license or higher for the users in scope.
Check for the Microsoft-managed policy. Microsoft’s managed policies page lists one named "Block device code flow." Microsoft creates these policies in report-only mode. It says it turns them on "no less than 30 days" after they appear in your tenant, with notice two weeks before. Look in Entra ID > Conditional Access > Policies for policies where Created by shows Microsoft. You can switch it on now instead of waiting.
Or build your own. Microsoft’s guide to blocking authentication flows gives these steps:
- Go to Entra ID > Conditional Access > Policies > New policy.
- Under users, include all users. Exclude your emergency access (break-glass) accounts.
- Under target resources, include all resources.
- Under Conditions > Authentication flows, set Configure to Yes and select Device code flow.
- Under Grant, select Block access.
- Set Enable policy to Report-only and create it.
- Review the results, then switch the policy to On.
Microsoft’s guidance is to get "as close as possible to a unilateral block on device code flow." Allow it only for documented cases, such as legacy tools that cannot be updated.
Find out who uses device code flow today
Before you turn the block on, check who would be affected. Microsoft’s documentation says to filter the Entra sign-in logs with the Authentication protocol filter set to device code. Report-only mode shows the same thing from the policy’s side.
The most common legitimate use in small offices is Microsoft Teams Rooms and other shared Teams devices. Microsoft has a separate guide for Teams devices. It says to put only the room accounts in an exception group and exclude that group. It warns that an exception "is account-scoped," so an excluded account can use device code flow for anything. Keep that group small and review it.
If you do not have Entra ID P1, you cannot build this policy. Staff guidance and fast session revocation become your main defenses, and the license is worth pricing.
What Google Workspace admins can do
Google’s 2-Step Verification deployment guide covers the strongest setting. In the Admin console, go to Security > Authentication > 2-step verification. Under enforcement methods, choose Only security key.
Google says that option "now supports both security keys and passkeys." It also says "Passkeys and security keys both have the same level of phishing protection." Start with admins and staff who handle money or member data. You can apply the setting to an organizational unit or group first.
Google’s admin account best practices add more steps. Give each administrator their own admin account, and give super admins a separate account for everyday work. Google also suggests enrolling super admins in the Advanced Protection Program.
If someone already signed in through a fake event page
A password reset alone may leave the attacker’s session working. End the sessions first.
Microsoft 365. Microsoft’s guide to revoking user access in an emergency says to open the user in Entra ID > Users, disable the account if needed, and select Revoke sessions. The PowerShell cmdlet is Revoke-MgUserSignInSession. Microsoft notes that access tokens last one hour by default, so some access can continue briefly.
Google Workspace. Google’s guide to managing a user’s security settings says to open the user under Directory > Users, click Security, and reset Sign-in cookies. Google says this signs the user out of their Google Account, and it "can take up to an hour" to end current Gmail sessions.
Then reset the password, check mail forwarding rules and recent sign-ins, and follow your data breach response plan if member or donor data was in the mailbox.
What to tell staff
Keep the message short:
- Never enter a sign-in code someone else gave you. A real device code only comes from a device in front of you, such as a TV or a meeting room screen.
- Free AI subscriptions in event invites are a red flag. Real event sign-ups rarely need your work account.
- If a sign-in page opened from an invite, close it. Go to the event organizer’s own site instead.
- Report it, even after the fact. Signing in by mistake is fixable if IT hears about it the same day.
Device code phishing is new to most staff, so say it plainly once. Our posts on phishing attacks and social engineering cover the wider training points.
Frequently Asked Questions
What is device code phishing?
It is an attack that abuses Microsoft’s device code sign-in. The attacker starts a sign-in, gives you the code, and you enter it on Microsoft’s real page. Microsoft then gives access to the attacker’s session.
Who found this campaign?
Push Security published its analysis on November 5, 2026. It says it detected the fake OpenAI, SpaceX and Anthropic event pages in late October 2026.
Does MFA stop device code phishing?
No. The victim completes their normal MFA on Microsoft’s real page, so the attacker gets a valid session. Blocking device code flow is the fix.
Was the Google path device code phishing too?
No. Push says the Google path used VoidProxy, a proxy that relays Google’s real sign-in and steals the session cookie. Passkeys and security keys defend against that.
How do I block device code flow in Microsoft 365?
Create a Conditional Access policy that targets the Device code flow authentication flow and blocks access. Start in report-only mode. Microsoft also offers a managed “Block device code flow” policy.
What license do I need?
Microsoft says Conditional Access needs Microsoft Entra ID P1 or higher for the users in scope.
Will blocking device code flow break anything?
It can break shared devices such as Microsoft Teams Rooms. Check sign-in logs first, and exclude only those device accounts, as Microsoft’s Teams devices guide describes.
What should we do if someone already signed in?
Revoke their sessions in Entra or reset sign-in cookies in Google Workspace, then reset the password. Check for new mail forwarding rules and unusual sign-ins.
Should we block the domains Push listed?
It does no harm, but Push says the operator rotates domains every few weeks. Use the list to search logs for past visits, and rely on the policy changes above.