SECURITY INCIDENT REPORT
Suspected OpenAI Brand Impersonation, Facebook Session Manipulation, and Potential Credential Exposure Through Apple TestFlight
Incident ID: IR-2026-10-26-ADSGPT Report Date: October 26, 2026 Classification: Security Research / Suspected Brand Impersonation / Potential Credential and Session Exposure Severity: HIGH — Provisional Confidence: High for application identification and observed code artifacts; undetermined for credential theft or exfiltration Status: Under Investigation Distribution: Apple Security, OpenAI Security, Meta Security, Google Firebase Abuse & Security
Reporting Organization: neurons.me Contact: media@neurons.me Affected Environment: Apple macOS 26.5.2, Apple Silicon (ARM64)
⸻
- EXECUTIVE SUMMARY
On October 26, 2026, an unsolicited Apple TestFlight invitation was received at media@neurons.me, advertising an application named OpenAI AdsGPT Campaign АI.
The invitation was distributed through Apple’s TestFlight infrastructure and identified the developer as:
AdsGPT Campaign Paid Media Platform OpCos, LLC
The application prominently used OpenAI branding, including the OpenAI logo, and advertised itself as:
Introducing OpenAI AdsGPT — GPT-6 Astra AI Campaign Ads Manager Pro
No evidence establishing an official relationship between the identified developer and OpenAI was available to the reporting party.
The application was installed on an Apple Silicon Mac and subsequently requested Facebook authentication.
Because of concerns regarding the application’s identity and authentication behavior, the application was removed from active use and a local forensic examination was initiated.
Static examination of the recovered application executable revealed code and diagnostic strings associated with:
- Facebook session-cookie restoration.
- JavaScript-based extraction of Facebook display names.
- Authenticated requests to Facebook web endpoints.
- Interaction with Facebook’s GraphQL endpoint.
- WebKit-based Facebook navigation.
- Firebase Firestore integration.
- Browser storage and session-management functionality.
These findings raise legitimate questions about whether the application implements an authorized Meta authentication flow or attempts to access Facebook session information through embedded web-browser functionality.
At the time of this report, no confirmed credential theft, session-token exfiltration, unauthorized account access, or persistent macOS compromise has been established.
Nevertheless, the combination of apparent OpenAI brand impersonation, unsolicited distribution, and Facebook session-handling functionality warrants prompt investigation by Apple, OpenAI, and Meta.
⸻
- INCIDENT TIMELINE
All times are local to the reporting environment, UTC−06:00, unless otherwise stated.
Date / Time Event October 26, 2026, 13:18 TestFlight invitation received October 26, 2026 Application installed through TestFlight October 26, 2026 Application opened and Facebook authentication requested October 26, 2026 User identified inconsistencies in developer identity and OpenAI branding October 26, 2026, approximately 13:53 Application moved to macOS Trash October 26, 2026 Static forensic examination initiated October 26, 2026 Application bundle identifier identified October 26, 2026 Apple TestFlight signing information examined October 26, 2026 WebKit storage and Firebase-related artifacts identified October 26, 2026 Facebook session-handling strings identified in executable October 26, 2026 Facebook password changed and existing sessions terminated October 26, 2026 Meta Business account review initiated
The precise application installation and execution timestamps should be confirmed against preserved system logs.
⸻
- DISTRIBUTION AND EMAIL METADATA
3.1 Invitation Details
Displayed Sender:
AdsGPT Campaign Paid Media Platform OpCos, LLC via TestFlight
Sender Email:
testflight_no_reply@email.apple.com
Recipient:
Date:
October 26, 2026
Time:
13:18
Subject:
AdsGPT Campaign Paid Media Platform OpCos, LLC has invited you to test OpenAI AdsGPT Campaign АI..
Mail Infrastructure:
Attribute Value From address testflight_no_reply@email.apple.com Mailed-by domain email.apple.com Signed-by domain email.apple.com Transport security TLS, as displayed by Gmail Distribution mechanism Apple TestFlight Named developer AdsGPT Campaign Paid Media Platform OpCos, LLC
The displayed Gmail authentication indicators are consistent with legitimate Apple email infrastructure.
However, the presence of a genuine TestFlight invitation does not establish the legitimacy of the application developer’s claimed affiliation with OpenAI.
The original RFC 5322 email headers have not yet been preserved or independently examined. Consequently, the complete SPF, DKIM, DMARC, and message-routing results remain outside the current forensic record.
3.2 Advertising Claims
The invitation included the following statements:
Limited Beta Access — 1,000 Users Worldwide
Selected participants may receive up to $200 in advertising credits.
The application also claimed to support:
- Campaign performance analytics.
- CTR, CPC, CPM, CPA, and ROAS analysis.
- Advertising account comparisons.
- Audience and creative evaluation.
- Campaign optimization recommendations.
- Advertising profitability analysis.
These claims could encourage users to connect valuable advertising accounts or disclose account credentials.
The promised advertising credits and limited-access framing should be examined as potential user-acquisition incentives.
⸻
- APPLICATION IDENTITY
4.1 Identified Application Metadata
Field Observed Value Advertised product OpenAI AdsGPT Campaign АI. Display name AdsGPT Developer AdsGPT Campaign Paid Media Platform OpCos, LLC Internal application name DragonTF Executable DragonTF Bundle identifier com.nwxb.pinsetter Application version 1.0 Build number 2 Developer Team ID 6Z7X992QYC Distribution Apple TestFlight Executable format Mach-O 64-bit CPU architecture ARM64 Original target platform iOS SDK reported in application metadata iPhoneOS 26.2 Xcode version reported in application metadata 26.3
4.2 Application Bundle Locations
Original application path:
/Applications/AdsGPT.app
Recovered application path:
/Users/suign/.Trash/AdsGPT.app
Embedded executable:
AdsGPT.app/Wrapper/DragonTF.app/DragonTF
Application container:
~/Library/Containers/com.nwxb.pinsetter/
The internal identifiers DragonTF and com.nwxb.pinsetter do not correspond directly to the advertised OpenAI AdsGPT product name.
Such differences are technically permissible and are not independently evidence of malicious activity.
However, they are important investigative identifiers.
4.3 Signing Information
The examined application reported:
Team Identifier:
6Z7X992QYC
Signing Authority:
TestFlight Beta Distribution
Certificate Chain:
- Apple Worldwide Developer Relations Certification Authority
- Apple Root CA
The application also contained the entitlement:
get-task-allow = false
These observations are consistent with a signed TestFlight-distributed application.
A prior strict code-signature verification produced the message:
resource envelope is obsolete (custom omit rules)
This warning requires additional interpretation and should not be treated as proof of tampering.
Investigative request to Apple: Determine the legal developer account associated with Team ID 6Z7X992QYC, confirm the app’s submission history, and investigate whether the use of OpenAI branding was authorized.
⸻
- APPLICATION ARCHITECTURE AND DEPENDENCIES
Static analysis identified an ARM64 Mach-O executable using Apple system frameworks and embedded references to Firebase components.
5.1 Apple Framework Dependencies
The executable dynamically references:
Framework / Library General Function UIKit.framework iOS user interface SwiftUI.framework Declarative user interface WebKit.framework Embedded web content Network.framework Network communication CFNetwork.framework Network services Security.framework Security APIs Foundation.framework Application services CoreFoundation.framework Core runtime services Combine.framework Reactive event handling SystemConfiguration.framework Network/system configuration CoreTelephony.framework Telephony-related APIs CoreGraphics.framework Graphics libc++.1.dylib C++ runtime libz.1.dylib Compression libobjc.A.dylib Objective-C runtime libSystem.B.dylib System runtime libswiftCore.dylib Swift runtime
Additional Swift runtime dependencies were also observed.
These libraries are legitimate Apple components.
Their presence does not establish malicious behavior.
5.2 Firebase Components
Static strings revealed references to:
- FirebaseFirestore
- FirebaseFirestoreInternal
- FirebaseCoreInternal
- FirebaseSharedSwift
- Firebase Remote Config components
Local application storage also contained Firestore-related artifacts.
An observed Firestore storage path included:
__FIRAPP_DEFAULT/video-51e24/main/000003.log
Potential Firebase Project Identifier:
video-51e24
This identifier should be treated as a lead requiring confirmation against the application’s Firebase configuration.
The presence of Firebase does not demonstrate data exfiltration.
However, the relevant Firebase project may provide investigators with information about application configuration, backend access, and the developer’s infrastructure.
⸻
- FACEBOOK SESSION-HANDLING FINDINGS
This section documents the most security-relevant findings from static examination.
6.1 Session-Cookie Restoration
The executable contained the string:
Restored previous Facebook session cookies.
Security Significance
This indicates the presence of code or diagnostic logic associated with restoring Facebook session cookies.
Potential legitimate explanations include maintaining an embedded browser session.
Potential security concerns include unauthorized session reuse or handling of authentication material outside an approved authorization workflow.
The static string alone does not establish whether a restoration operation actually occurred during the observed execution.
Finding Classification: Suspicious session-management capability.
Confidence: High that the string exists in the executable; undetermined whether the behavior executed.
⸻
6.2 JavaScript-Based Facebook Identity Extraction
The executable contained:
Successfully extracted Facebook Display Name (JS):
And:
Successfully extracted Facebook Display Name (navigation):
Additional diagnostic text referenced:
Direct JS objects in Facebook JS runtime
The application also contained logic referencing Facebook mobile interfaces:
Security Significance
These strings indicate functionality designed to obtain Facebook display-name information through web navigation or JavaScript.
A display name is not equivalent to an authentication credential.
Nevertheless, these mechanisms merit examination because they suggest the application interacts directly with Facebook’s web interface rather than relying exclusively on a conventional authorization API.
Finding Classification: Web-interface identity extraction.
Confidence: High for code presence; execution and authorization status unconfirmed.
⸻
6.3 Credentialed Facebook Requests
The executable contained the JavaScript expression:
fetch('https://www.facebook.com/', { credentials: 'include' }).catch(() => {});
Security Significance
The credentials: 'include' option instructs the browser to include applicable credentials, including cookies, subject to browser security controls.
This is not inherently malicious.
However, in combination with session-cookie restoration functionality, it warrants investigation into how the application handles authenticated Facebook sessions.
Finding Classification: Credentialed web request capability.
⸻
6.4 Facebook GraphQL Interaction
The executable contained a reference to:
https://www.facebook.com/api/graphql/
The surrounding code included a JavaScript fetch() invocation.
Security Significance
Facebook’s web interface uses GraphQL-based endpoints for application functionality.
Direct interaction with such endpoints may be legitimate in some contexts, but it should be examined to determine whether the application is using supported authentication mechanisms and authorized access.
The current evidence does not establish which GraphQL operations were executed or whether account information was extracted.
Finding Classification: Direct Facebook web API interaction.
⸻
6.5 Browser Cookie Manipulation
The executable included JavaScript assigning:
m_pixel_ratio
And:
wd
as cookies under the .facebook.com domain.
These cookies are associated with browser display characteristics.
They are not, by themselves, Facebook authentication cookies.
Their presence demonstrates cookie-management code but does not establish session hijacking.
⸻
- WEBKIT FORENSIC ARTIFACTS
Local application storage included WebKit-related directories and databases.
Observed components included:
- WebsiteData
- NetworkCache
- LocalStorage
- HTTPStorages
Relevant database filenames included:
Database Investigative Relevance observations.db Domain observations and relationship records localstorage.sqlite3 Web local-storage metadata pcm.db Private click-measurement data ResourceMonitorPersistence.db Resource-monitoring records AlternativeService.sqlite Alternative network-service metadata httpstorages.sqlite HTTP storage metadata
7.1 Observed Domains
The WebKit observation database contained references to:
These domains are associated with Meta infrastructure.
7.2 Redirect-Related Records
The SubresourceUniqueRedirectsTo table contained:
1|1793043965.09444|2 2|1793043965.09508|3
The records require interpretation according to the exact database schema and foreign-key relationships.
The domain identifiers previously observed suggest relationships among Facebook-related domains.
However, this database does not provide a complete network-traffic history.
No confirmed exfiltration destination was identified from the examined WebKit domain records.
This does not exclude network communication through other application components.
⸻
- POTENTIAL SECURITY RISKS
The following scenarios warrant investigation.
8.1 OpenAI Brand Impersonation
Risk: HIGH
The application uses OpenAI’s name and visual identity while identifying a different company as the developer.
The legitimacy of this use has not been established.
Potential impact includes user deception and unauthorized collection of account access.
8.2 Facebook Session Exposure
Risk: HIGH — Conditional
The executable contains functionality associated with Facebook cookies, JavaScript interaction, and authenticated web requests.
Potential impact depends on whether authentication material was accessed, stored, transmitted, or reused.
No such compromise has been confirmed.
8.3 Advertising Account Compromise
Risk: HIGH — Conditional
The application specifically targets advertising workflows.
An unauthorized integration with Meta Business could potentially expose advertising data or permit account modifications, depending on the permissions granted.
No unauthorized advertising-account changes have been established.
8.4 Unauthorized Backend Collection
Risk: UNDETERMINED
Firebase components were identified.
Further examination is required to determine what application data, if any, was transmitted to Firebase services.
No confirmed credential exfiltration to Firebase has been observed.
8.5 Persistent Device Compromise
Risk: NOT ESTABLISHED
Initial checks did not identify obvious persistence mechanisms associated with the application’s advertised name.
These checks were not exhaustive.
There is currently insufficient evidence to classify the application as persistent macOS malware.
⸻
- USER CONTAINMENT ACTIONS
The reporting party has taken the following protective actions:
Action Status Application removed from active use Completed Facebook password changed Completed Existing Facebook sessions terminated Completed Two-factor authentication Previously enabled Meta Business access review In progress Original application bundle preserved in Trash Reported Application container examined Completed in part Static executable analysis Completed in part Full network-traffic reconstruction Pending Independent malware analysis Pending
These actions reduce the immediate risk associated with potentially exposed Facebook credentials.
Further review of Meta Business administrators, partners, connected applications, and advertising-account permissions is recommended.
⸻
- IDENTIFIERS FOR INVESTIGATORS
The following identifiers may assist in correlating records across Apple, OpenAI, Meta, and Google systems.
Category Identifier Advertised product OpenAI AdsGPT Campaign АI. Display name AdsGPT Developer legal name displayed AdsGPT Campaign Paid Media Platform OpCos, LLC Application bundle ID com.nwxb.pinsetter Internal application name DragonTF Executable DragonTF Apple Developer Team ID 6Z7X992QYC Application version 1.0 Build 2 Distribution channel Apple TestFlight Invitation sender testflight_no_reply@email.apple.com Invitation domain email.apple.com Recipient media@neurons.me Potential Firebase project ID video-51e24 Target architecture ARM64 Executable format Mach-O 64-bit Original target OS iOS Host OS macOS 26.5.2 Host build 25F84 Facebook domain facebook.com Facebook CDN fbcdn.net Facebook supporting domain fbsbx.com Facebook mobile endpoint m.facebook.com Facebook basic mobile endpoint mbasic.facebook.com Facebook GraphQL endpoint www.facebook.com/api/graphql/
Evidence Not Yet Collected
The following items should be obtained before making definitive claims about malicious activity:
- Full original email headers.
- Complete application bundle SHA-256 hashes.
- Apple developer certificate serial number.
- Full provisioning profile metadata, where available.
- Complete executable entitlements.
- Firebase application configuration.
- Confirmed application network destinations.
- Network request and response evidence.
- Evidence of actual cookie access or transmission.
- Facebook authorization and permission records.
- Relevant Meta account security events.
- Complete forensic acquisition timestamps.
⸻
- REQUESTED ACTIONS
11.1 Apple Security and App Review
We request that Apple:
- Investigate the developer account associated with Team ID 6Z7X992QYC.
- Verify whether the application is authorized to use OpenAI branding.
- Review the application’s TestFlight submission and distribution history.
- Examine its Facebook authentication and cookie-handling implementation.
- Determine whether the application violates TestFlight or App Store policies.
- Preserve relevant developer-account and application-distribution records.
- Consider suspending distribution if policy violations or user-security risks are confirmed.
- Investigate whether related applications have been distributed under other developer accounts.
11.2 OpenAI Security and Brand Protection
We request that OpenAI:
- Confirm whether the application is an authorized OpenAI product or partnership.
- Investigate potential unauthorized use of the OpenAI name and logo.
- Review the product’s claims involving GPT-6 and Astra.
- Coordinate with Apple regarding potential brand impersonation.
- Consider appropriate brand-protection or abuse-reporting action.
11.3 Meta Security
We request that Meta:
- Review whether the application implements an authorized Facebook authentication flow.
- Investigate the application’s Facebook cookie-restoration functionality.
- Examine whether its GraphQL interactions violate platform security requirements.
- Assess whether any unauthorized session access can be correlated with the incident.
- Investigate any associated application integrations or advertising-account permissions, where legally and technically available.
11.4 Google Firebase Security
We request that Google:
- Determine whether video-51e24 corresponds to the application under investigation.
- Review the project’s association with the identified developer, subject to applicable legal and privacy requirements.
- Assess whether the project has been reported for abusive activity.
- Preserve relevant evidence if abuse is substantiated.
⸻
- EVIDENCE PRESERVATION
The reporting party has retained or identified the following evidence sources:
Evidence A — Original TestFlight Invitation
Includes developer identity, advertised product name, recipient, timestamp, and marketing claims.
Evidence B — TestFlight Application Screenshot
Shows the application’s presentation within Apple’s beta distribution interface.
Evidence C — Recovered Application Bundle
Path:
~/.Trash/AdsGPT.app
Evidence D — Application Container
Identifier:
com.nwxb.pinsetter
Evidence E — Static Binary Analysis
Includes file, otool, and strings output from the ARM64 executable.
Evidence F — WebKit Databases
Includes local browser observation and storage artifacts.
Evidence G — System Logs
Includes application-related installation, execution, and removal events previously examined.
Preservation Recommendations
Before additional modification or deletion:
- Calculate SHA-256 hashes of preserved evidence.
- Record acquisition times.
- Preserve original files and analyze copies.
- Retain the original email in an exportable format.
- Avoid executing the recovered application.
- Store sensitive session material securely.
- Do not publish live cookies, access tokens, passwords, or private account identifiers.
⸻
- ASSESSMENT AND CONCLUSION
The investigation identified an unsolicited TestFlight application advertising itself using OpenAI branding while being distributed under a separately named developer account.
The recovered application contains functionality associated with Facebook web authentication, cookie restoration, JavaScript-based identity extraction, and authenticated requests to Facebook web services.
These characteristics are security-relevant because the application specifically targets users managing advertising accounts.
The observed behavior is sufficient to justify investigation by the platform providers.
However, the current forensic record does not establish that the application successfully stole credentials, exfiltrated session cookies, compromised advertising accounts, or installed persistent malware.
The principal confirmed concerns are:
- Potentially unauthorized OpenAI brand usage.
- Facebook session-handling functionality requiring security review.
- Unsolicited distribution of an application requesting access to valuable third-party accounts.
- An identifiable Apple developer account and application bundle suitable for platform-level investigation.
Recommended Disposition
Escalate for formal security and abuse review.
Request that Apple and OpenAI verify the developer’s authorization to distribute the application under OpenAI branding.
Request that Meta review the authentication and session-handling implementation.
If unauthorized branding, deceptive authentication, or credential abuse is established, request appropriate enforcement action, including removal from distribution and investigation of related applications.
⸻
FINAL INCIDENT CLASSIFICATION
Suspected Brand Impersonation and Potential Facebook Session Security Risk
Severity: HIGH — Provisional
Evidence Status: Partial forensic evidence; static analysis completed in part.
Confirmed Credential Theft: No.
Confirmed Exfiltration: No.
Confirmed Persistent Malware: No.
Platform Investigation Recommended: Yes.
⸻
Prepared for security investigation and responsible disclosure.
Reporting Organization: neurons.me Incident Date: October 26, 2026 Reference: IR-2026-10-26-ADSGPT
End of Report
About the author
- PoC || GTFO
Comments
Join the discussionNo comments yet. Be the first to comment.
