In a context where artificial intelligence has established itself as a daily tool, the security of exchanges with these technologies becomes a major concern. In 2025, a shocking revelation shook the digital world: millions of private conversations with ChatGPT were hacked and put up for sale on the black market. This massive data leak raises crucial questions about privacy, cybersecurity, and the risks faced by users of these AI platforms. Behind this breach, a widely adopted Chrome extension recommended by Google, Urban VPN Proxy, was at the origin of the illegal and systematic capture of these private exchanges.
This extension, praised for its simple features and reassuring interface, benefited from increased trust thanks to an official badge awarded by Google, prompting more than six million internet users to use it without suspicion. Yet, unbeknownst to them, these users were exposing sensitive information: medical questions, financial data, and even professional secrets were sucked up and redistributed to third parties. This large-scale scandal highlights a new form of sophisticated and meticulously orchestrated cyberattack, which exploits users’ naivety and the vulnerabilities of trusted platforms to orchestrate an illegal sale of personal data.
- 1 Urban VPN Proxy: a hijacked extension causing a massive leak of private ChatGPT conversations
- 2 From data collection to marketing analysis: diverting private conversations into commercial currency
- 3 The concealment behind the privacy policy: Urban Cyber Security Inc.’s double talk
- 4 Timeline and recommendations: what you need to know about the ChatGPT data leak since July 2025
- 5 What to do in the face of this massive private conversation hack and how to better protect your privacy in the AI era?
Urban VPN Proxy: a hijacked extension causing a massive leak of private ChatGPT conversations
Urban VPN Proxy, presented as a legitimate tool for protecting online privacy, quickly won a large audience due to its ease of use and increased visibility, notably through an official recommendation badge granted by Google. This recognition amplified its legitimacy among internet users, who saw it as a reliable solution against online threats. Unfortunately, this trust proved largely misplaced.
An in-depth investigation conducted by the specialized firm Koi revealed that the extension embedded scripts invisible to the user. These scripts continuously collected all conversations held via various AIs, including ChatGPT, but also Gemini, Claude, Grok, and DeepSeek. Contrary to what might seem obvious, the actual activation of the VPN had no impact on this fraudulent collection: from the moment of installation, every exchange became accessible to third parties without any alert or clear consent.
This sophisticated interception method constituted a sneaky form of cyberattack that literally violated the confidentiality of personal data exchanged in these supposedly protected spaces. Thus, private dialogues, containing personal, professional, or even sensitive information, were sucked up and stored without users having the slightest idea.
The stakes of this hacking go beyond a simple technical abuse: this intrusion directly affects the trust between users and artificial intelligence tools. How can data continue to be entrusted when confidentiality is compromised from the very first connection? The consequences affect both individuals and businesses, becoming an emblematic case illustrating the urgent need for strengthened cybersecurity.
From data collection to marketing analysis: diverting private conversations into commercial currency
The quantity and nature of conversations intercepted by the extension go far beyond simple anonymous exchanges. Koi confirmed that the collected data included numerous sensitive contents: health-related questions, financial information, discussions related to professional computer code, as well as deeply intimate personal dilemmas.
This diversity illustrates the richness of the volume of stolen information, which then serves to feed marketing analyses. The actors behind this operation exploited these exchanges by monetizing them with commercial partners via an affiliated company, BiScience. This company harvested the data to profit from it, thus turning the data leak into a real economic lever.
The deployed economic model relies on the illegal resale of data to marketing entities that take advantage of users’ behaviors, needs, and preferences based on these compromised conversations. This reality raises a troubling paradox: the use of AI tools, meant to improve our daily lives, can lead to unregulated and opaque exploitation of our private lives.
No option in the extension’s settings allowed blocking this collection, making the removal of the extension the only effective measure to stop the leak. Forbes corroborated these facts, confirming that even savvy users had no real control. This situation highlights the fragility of current protection mechanisms, facing increasingly sophisticated data surveillance and extraction techniques.
List: Types of private data exposed through Urban VPN Proxy
- Medical questions — diagnoses, treatments, and personal advice
- Financial information — bank details, wealth management strategies, transactions
- Professional codes — program fragments, scripts, sensitive instructions
- Personal dilemmas — confidences, intimate decisions, family exchanges
- Strategic discussions — business plans, internal projects, key figures
The concealment behind the privacy policy: Urban Cyber Security Inc.’s double talk
The publishing company, Urban Cyber Security Inc., publicly assumed the data collection practice through its privacy policy, which mentioned sharing browsing data with a partner company named BiScience. However, this disclosure remained very partial, and the official discourse, notably on the Chrome Web Store page, explicitly denied any resale to third parties, claiming to limit the use of data to the extension’s main functionality.
This contradiction raises serious doubts about the company’s transparency and ethics. Users were subjected to ambiguous communication that led them to believe in guaranteed security and confidentiality, while in reality their personal conversations were being illegally sold without anyone’s knowledge.
The scandal does not stop there: the investigation revealed that seven other extensions from the same publisher used the same data extraction mechanism. These extensions had more than two million active users combined and all bore a recommendation badge that could suggest reliable and objective official validation.
This system reveals a multi-level exploitation model, consciously implemented by the publisher to maximize access to extremely lucrative data, under a veneer of legitimacy. The cybersecurity stakes in this context are crucial, as it involves protection against cyberattack risks that will exploit this data for malicious or commercial purposes.
Timeline and recommendations: what you need to know about the ChatGPT data leak since July 2025
The intercepted conversations date back to July 2025 and include every exchange made with an AI since that date, without any distinction. This massive and prolonged backup increases the seriousness of the leak: this is not an isolated incident, but rather continuous collection over several months. The data in question was transmitted to commercial partners, suggesting wide and uncontrolled international dissemination.
Idan Dardikman, a renowned cybersecurity researcher, issued a clear and urgent warning to all internet users: any extension created by Urban Cyber Security Inc. must be removed immediately. He also recommends carefully rereading the privacy policies of the applications and extensions used in order to detect any non-compliant data sharing clauses.
This case sets a textbook example of the risks associated with massive absorption of personal data, exposing the vulnerabilities of mainstream platforms against growing and sophisticated cyberattacks. It also demonstrates the necessity for internet users to adopt increased vigilance and to favor solutions reputed to be secure and transparent.
| Key Dates | Major Events | Consequences |
|---|---|---|
| July 2025 | Start of systematic collection of private conversations via Urban VPN Proxy | Continuous exposure of sensitive data to unauthorized third parties |
| October 2025 | Publication of Koi’s investigation revealing automatic collection | Media awareness and alert to users |
| November 2025 | Independent verifications and confirmation by Forbes | Strengthening criticism against Urban Cyber Security Inc. |
| December 2025 | Warning by cybersecurity researchers and recommended removal of extensions | Call for vigilance and corrective measures among users |
What to do in the face of this massive private conversation hack and how to better protect your privacy in the AI era?
The leak revealed by this case highlights the urgent need for better management of data privacy in an expanding digital environment. The multiplication of cyberattacks targeting private conversations, especially with AIs like ChatGPT, requires heightened awareness from users.
To effectively guard against this type of cyberattack, it is crucial:
- To rigorously verify the origin and reputation of extensions or applications before installation.
- To avoid recording or sharing sensitive information on platforms whose privacy policies are not mastered.
- To immediately delete any suspicious extension or those linked to dubious publishers.
- To use established security solutions, combined with good antivirus software and regular device updates.
- To carefully read privacy policies and terms of use, taking into account community comments and alerts.
Aware that this violation of the private sphere can have lasting impacts, some experts also advocate for strengthening legislation to more strictly regulate the collection and commercialization of personal data. Better regulation would protect the fundamental rights of internet users and deter these abusive practices.
This increased vigilance remains an essential shield against the growing sophistication of attacks, but also a collective responsibility in a constantly evolving digital world.
