Threat reportPhishingTL-2026-2548
Fake ChatGPT Billing Email Phishing Campaign Abuses Google API Redirect to Steal OpenAI Credentials via nxcli.io
Cofense's Phishing Defense Center identified an active credential-phishing campaign impersonating OpenAI with a fake ChatGPT billing notice demanding payment of a fabricated $23.80 balance within 48 hours. The email's payment button routes victims through a legitimate Google API redirect (notifications.googleapis.com) to a counterfeit ChatGPT login page on an abused nxcli.io hosting subdomain that harvests OpenAI account credentials.
Key facts
| Identifiers | TL-2026-2548 |
|---|---|
| Severity | MEDIUM |
| Affected products | OpenAI ChatGPT / OpenAI Platform account holders (individual and business billing contacts) |
| MITRE ATT&CK techniques | T1036.005 Match Legitimate Resource Name or Location, T1111 Multi-Factor Authentication Interception, T1204.001 Malicious Link, T1566.002 Spearphishing Link, T1583.006 Web Services, T1585.002 Email Accounts, T1586.002 Email Accounts, T1598.003 Spearphishing Link, T1608.005 Link Target, T1657 Financial Theft, T1684.001 Impersonation |
| Motivation | FINANCIAL |
| Target sectors | technology, cross-sector |
| Target regions | Global |
| Detection rules | 9 (1 shown free) |
| Sources | 8 cited sources |
Detection coverage
As of 2026-09-17, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-2548 across Splunk SPL, Microsoft KQL and Sigma, covering 15 indicator(s) of compromise. The whole corpus is readable without an account. Every threat comes with one complete detection in Splunk SPL, Microsoft KQL and Sigma, free for everyone; a free account adds daily TL Creds to unlock the rest. Paid plans add a larger daily allowance, correlation and the MCP server. Threadlinqs MCP server · View plans.
Free detection: Fake ChatGPT Billing Email — Spoofed Sender + Google-Redirect Lure
Detects inbound email matching the Cofense-documented fake ChatGPT/OpenAI billing lure: sender on the nxcli.io shared-hosting domain (or the corroborating bellnet.ca compromised-webmail wave), a payment-urgency subject line, and a link whose first hop is notifications.googleapis.com before redirecting off-Google. Catches the campaign at initial delivery, before any user click.
- Languages: Splunk SPL, Microsoft KQL, Sigma
- ATT&CK: T1566.002, T1585.002, T1586.002
- Data sources: m365_defender_email, proofpoint_pod, email_gateway_logs
- Severity: high
- Confidence: high
Splunk SPL
index=email sourcetype="ms365:defender:email" OR sourcetype="proofpoint:pod:message"
| eval sender_domain=lower(mvindex(split(sender_address, "@"), -1))
| eval is_susp_sender=if(match(sender_domain, "nxcli\.io$") OR sender_domain="bellnet.ca", 1, 0)
| eval is_billing_lure=if(match(lower(subject), "update your payment method|payment method to avoid|service interruption|chatgpt|openai"), 1, 0)
| eval has_google_redirect=if(match(url_list, "notifications\.googleapis\.com"), 1, 0)
| where (is_susp_sender=1 AND is_billing_lure=1) OR (is_billing_lure=1 AND has_google_redirect=1)
| table _time, message_id, recipient, sender_address, sender_domain, subject, url_list, spf_result, dkim_result
| sort - _timeMicrosoft KQL
EmailEvents
| where Timestamp > ago(24h)
| extend SenderDomain = tolower(tostring(split(SenderFromAddress, "@")[1]))
| where SenderDomain endswith "nxcli.io" or SenderDomain == "bellnet.ca"
| join kind=inner (
EmailUrlInfo
| where Timestamp > ago(24h)
| where Url has "notifications.googleapis.com" or Url has "nxcli.io"
) on NetworkMessageId
| where Subject has_any ("Update Your Payment Method", "Service Interruption", "ChatGPT", "OpenAI")
| project Timestamp, NetworkMessageId, SenderFromAddress, RecipientEmailAddress, Subject, Url, ThreatTypesSigma
title: Fake ChatGPT Billing Phishing Email Delivered via nxcli.io / Google Redirect
id: 7a1c9e2a-4f3b-4c8e-9a2d-1b6e8f0c3d11
status: experimental
description: Detects delivery of a fraudulent ChatGPT/OpenAI billing email chaining a notifications.googleapis.com redirect to an nxcli.io credential-phishing page, or sent from the corroborating bellnet.ca wave.
references:
- https://www.helpnetsecurity.com/2026/09/17/chatgpt-phishing-email-openai-password/
- https://malwaretips.com/blogs/chatgpt-payment-method-update-email-scam/
author: Threadlinqs Team
date: 2026-09-17
tags:
- attack.initial-access
- attack.t1566.002
- attack.resource-development
- attack.t1585.002
- attack.t1586.002
- attack.tool.phishing_kit
logsource:
category: threat_detection
product: m365
service: threat_intel
detection:
selection_sender:
SenderFromAddress|endswith:
- '.nxcli.io'
SenderFromAddress|contains: 'bellnet.ca'
selection_lure:
Subject|contains:
- 'Update Your Payment Method'
- 'Service Interruption'
- 'ChatGPT'
- 'OpenAI'
selection_redirect:
Url|contains:
- 'notifications.googleapis.com'
- 'nxcli.io'
condition: (selection_sender and selection_lure) or (selection_lure and selection_redirect)
falsepositives:
- Legitimate OpenAI billing notifications sent from openai.com or stripe.com (verify SPF/DKIM alignment before acting)
- Internal security-awareness phishing simulations using similar templates
level: highKnown false positives:
- Legitimate OpenAI/Stripe billing mail that happens to reference payment urgency; verify sender domain and SPF/DKIM alignment before escalating
Tuning notes: Maintain the nxcli.io / bellnet.ca sender list as a living watchlist lookup rather than hardcoding, since MalwareTips confirms the same lure template is redistributed across rotating sending infrastructure; widen the subject regex cautiously to avoid false-flagging real OpenAI billing mail.
- M365 Defender: ChatGPT Billing Phish Delivery from [indicator] / [indicator] · unlock with TL Creds
- Sentinel: Endpoint Redirect Chain notifications.googleapis.com → [indicator] login.php/key.php · unlock with TL Creds
- Sentinel: High-Risk OpenAI/ChatGPT SSO Sign-In Following [indicator] Phishing-Page Visit · unlock with TL Creds
- Sigma: Fake ChatGPT Billing Phishing Email ([indicator] / [indicator] Sender, Google-Redirect Lure) · unlock with TL Creds
- Sigma: Google API Open-Redirect Chaining to [indicator] Cloned ChatGPT Login (login.php/key.php) · unlock with TL Creds
- Sigma: High-Risk OpenAI/ChatGPT Sign-In Correlated to Prior [indicator] Phishing-Page Visit · unlock with TL Creds
- User Click-Through: Google API Open-Redirect Chaining to [indicator] Fake ChatGPT Login · unlock with TL Creds
- Post-Click Account Compromise: Anomalous OpenAI/SSO Sign-In or Billing Change Following Phishing-Page Visit · unlock with TL Creds
Indicators of compromise
As of 2026-09-17, Threadlinqs Intelligence tracks 15 indicators of compromise for TL-2026-2548.
Sample indicators (3 of 15)
- Domain:
nxcli[.]io - URL:
hxxps://notifications[.]googleapis[.]com - Domain:
bellnet[.]ca
Defanged, TLP:CLEAR. Full set of 15 indicators: unlock with TL Creds.
Technical analysis
On 2026-09-17, Cofense Phishing Defense Center analyst Josh Varden publicly documented a mass credential-phishing campaign impersonating OpenAI/ChatGPT billing communications. The lure email, sent from support@9527db6e1a.nxcli.io (a subdomain of the shared hosting domain nxcli.io, not an OpenAI-owned domain), carries the subject 'Urgent: Update Your Payment Method to Avoid Service Interruption,' the ChatGPT logo, and a fabricated outstanding balance of $23.80, with a 48-hour countdown to service suspension designed to pressure the recipient into immediate, unconsidered action.
The email's 'Update Payment Information' button does not link directly to the attacker's infrastructure. Instead, its URL begins at notifications.googleapis.com, a legitimate Google API domain, which redirects the browser onward to the actual phishing payload hosted on the nxcli.io subdomain. This is a defense-evasion pattern Cofense has separately documented as a broader 2024-2025 trend: threat actors chain phishing links through trusted first-party cloud domains (Google AMP, google.com/url, storage.googleapis.com, and now notifications.googleapis.com) specifically because secure email gateways (SEGs) and reputation-based URL filters extend implicit trust to Google-owned domains and often only inspect the first hop, not the final redirect destination.
The landing page clones the ChatGPT/OpenAI sign-in experience, displaying the ChatGPT logo and a 'Welcome back' greeting above username and password fields, visually approximating the legitimate auth.openai.com login page. Cofense identified two paths on the phishing host: login.php, which captures the submitted username and password, and key.php, a second path suggesting a follow-on data-capture or verification step after initial credential entry. After submission, victims are redirected to a generic error page rather than back to a real OpenAI property, a common technique to reduce victim suspicion and avoid immediately confirming the compromise.
This campaign is not an isolated event. Check Point's Q2 2026 Brand Phishing Report found that ChatGPT entered the top 10 most-impersonated brands in phishing attacks for the first time, following a fraudulent 'ChatGPT Plus payment failed' email Check Point observed in June 2026 using the same billing-failure social-engineering pattern. A separate, independently documented wave of the identical $23.80 billing lure (invoice reference INV-2026-08-10-001, dated 2026-08-10) was reported by MalwareTips on 2026-08-13, using a different sending domain (bellnet.ca — a legitimate Canadian ISP webmail domain, per a corroborating OpenAI community forum report, indicating a likely compromised consumer email account rather than attacker-registered infrastructure) but the same fabricated amount, urgency framing, and harvesting workflow — indicating the underlying phishing kit or lure template is being reused and redistributed across multiple sending infrastructures rather than being unique to the nxcli.io wave. MalwareTips documents this corroborating wave's data-collection form as a nine-step sequence: after the ChatGPT/email credentials (email + password) are captured, the same fake flow proceeds to request the full payment card number, expiration date, CVV security code, cardholder name, billing address, and phone number, and then prompts for a bank-issued one-time verification code (OTP) before landing on a reassuring confirmation page designed to delay victim suspicion and detection. No CVE or software vulnerability is implicated; this is a pure social-engineering and infrastructure-abuse campaign targeting OpenAI/ChatGPT account credentials, payment-card data, and bank OTP codes.
MITRE ATT&CK techniques by tactic:
Defense Evasion
T1036.005 Match Legitimate Resource Name or Location; T1684.001 Impersonation
Credential Access
T1111 Multi-Factor Authentication Interception
Execution
Initial Access
Resource Development
T1583.006 Web Services; T1585.002 Email Accounts; T1586.002 Email Accounts; T1608.005 Link Target
Reconnaissance
Impact
Remediation
Immediate actions
- Block or quarantine inbound mail referencing the nxcli.io sending/hosting infrastructure (support@9527db6e1a.nxcli.io) and mail chaining through notifications.googleapis.com toward non-Google destinations paired with ChatGPT/OpenAI billing lures
- Instruct users who clicked the link and entered credentials to immediately reset their OpenAI/ChatGPT password and any reused passwords on other services
- Review OpenAI/ChatGPT account login and billing activity for signs of unauthorized access or payment-method changes following a suspected click
- Report confirmed phishing samples to OpenAI (help.openai.com) and, for US users, ReportFraud.ftc.gov
Workarounds
- Always navigate directly to chatgpt.com or type auth.openai.com rather than clicking links in billing/payment emails to manage ChatGPT subscription or payment settings
- Verify the browser address bar reads auth.openai.com (not a lookalike or third-party hosting domain) before entering any OpenAI credentials
- Treat unsolicited 'account suspension in 48 hours' urgency framing in billing emails as a phishing indicator regardless of the amount claimed
Longer-term hardening
- Extend security-awareness training to cover AI-platform billing/subscription phishing specifically, given ChatGPT's first-time entry into Check Point's top-10 most-impersonated-brands ranking in Q2 2026
- Enable and enforce multi-factor authentication on OpenAI/ChatGPT accounts, prioritizing Team/Enterprise admin and billing-owner accounts
- Tune secure email gateways and URL-reputation tooling to follow and inspect final redirect destinations rather than granting implicit trust to first-hop Google API/AMP/Cloud Storage domains, per Cofense's documented Google-redirect-abuse trend
- Establish a recurring watch for newly registered or newly observed subdomains on shared/free hosting platforms (e.g., nxcli.io) impersonating the organization's SaaS vendors
Affected and fixed versions
- OpenAI — ChatGPT / OpenAI Platform account holders (individual and business billing contacts)
Vulnerable versions: N/A — social-engineering campaign, not a software vulnerability; targets any OpenAI/ChatGPT account holder who receives the lure email
Fixed in: N/A
Timeline
- Cofense publishes broader research documenting threat actors chaining phishing links through legitimate Google-owned redirect endpoints (Google AMP, google.com/url, Google Translate, Google Maps, Google Storage) to bypass secure-email-gateway URL reputation checks — the same trusted-domain redirect-abuse technique later observed via notifications.googleapis.com in the locked-ioc-8 ChatGPT billing campaign
- Cofense publishes data showing malicious email attacks accelerated to one every 19 seconds in 2025 (from one every 42 seconds in 2024), the detection-volume backdrop behind the Phishing Defense Center's continued identification of brand-impersonation campaigns such as this one
- Check Point's Q2 2026 Brand Phishing Report is reported showing ChatGPT entering the top 10 most-impersonated brands for the first time, citing an earlier fraudulent 'ChatGPT Plus payment failed' email observed in June 2026 using the same billing-failure lure pattern
- A parallel wave of the identical $23.80 fake ChatGPT/OpenAI billing lure (invoice reference INV-2026-08-10-001) circulates using a different sending domain (locked-ioc-7), reusing the same urgency template and payment-harvesting workflow but distinct infrastructure from the locked-ioc-8 wave
- MalwareTips publishes a technical writeup of the related $23.80 ChatGPT payment-update phishing wave (invoice INV-2026-08-10-001), documenting a nine-step workflow that harvests ChatGPT email/password credentials, then full payment card number, expiration, CVV, cardholder name, billing address, and phone number, then a bank-issued one-time verification code, before showing a reassuring confirmation page
- Help Net Security publishes coverage of the Cofense analysis, detailing the $23.80 fake balance, 48-hour suspension threat, the sender address locked-ioc-4, and the Google-redirect defense-evasion chain
- Cofense Phishing Defense Center analyst Josh Varden identifies and traces the campaign's 'Update Payment Information' button through a notifications.googleapis.com redirect to the locked-ioc-8 credential-harvesting page (login.php, key.php)
Sources cited
- Which Brands Are Impersonated Most? Inside the Q2 2026 Brand Phishing Report Check Point · Vendor report · 2026-07-24
- Google Redirect Abuse in 2024: Key Trends & Tactics cofense.com · Vendor report · 2025-07-30
- Cofense Report Reveals AI-Powered Phishing Accelerated to One Attack Every 19 Seconds businesswire.com · Vendor report · 2026-02-04
- ChatGPT Payment Method Update Email Scam: Avoid This Fake Bill malwaretips.com · Blog · 2026-08-13
- Scam Alert: Fake OpenAI Subscription Email Requesting Payment Details community.openai.com · Blog · 2026-08-01
- A fake ChatGPT billing email is after your OpenAI password Help Net Security · News · 2026-09-17
- ChatGPT Among Top 10 Most Impersonated Brands in Phishing Attacks infosecurity-magazine.com · News · 2026-07-24
- ChatGPT joins the most impersonated brands in phishing attacks Help Net Security · News · 2026-07-27
Previous and next in phishing
Drafted by the Threadlinqs research pipeline from 8 sources; published ; last updated —.