Code out of access: Scammers started selling fake Claude accounts
Thousands of ads have appeared in the Russian Federation for the sale of accounts of the Claude neural network with a verified identity, as well as with offers to undergo document verification (KYC) for the user. At the same time, information about the owners of the documents used for verification often remains inaccessible to buyers. The average cost of such a service is about 4 thousand rubles. Demand for it increased after the developer of Anthropic tightened the rules for access to its products and began requesting identification documents from Russian users. However, most often they become victims of scams. In addition, an American company may re-check it or block it due to the owner's country of residence or place of work.
The market for other people's documents
Since November 12, Anthropic has tightened the rules for using its products. The new restrictions apply not only to users who are physically located in unsupported countries, but also to employees of organizations registered or headquartered in such regions. In particular, this applies to Russian companies, even if their employees work abroad. The requirements affect Claude, Claude Code, Cowork, and API. For the first time, Anthropic introduced restrictions for Russians in September 2025.
According to experts from MWS AI (part of MTS), since the beginning of the year, thousands of offers have appeared on domestic platforms for the sale of verified accounts of the Claude neural network and for help with KYC verification. Some intermediaries offer ready-made accounts, while others undertake to verify the identity directly in the customer's account. To do this, they use documents from countries where the service is officially available. At the same time, buyers are often not informed who exactly owns the documents used for verification.
Intermediaries charge about 4 thousand rubles for identity verification, the company added. At the same time, sellers warn that they do not guarantee that the account will be saved after repeated identity verification. If Anthropic requests the documents again, the buyer will have to solve this problem on their own. There are no guarantees against account blocking, service suspension, or payment problems. Identity verification also does not apply to the Claude API, through which developers connect the neural network to their own applications and services.
Cross Technologies recorded about 400 active ads for the sale of Claude accounts on thematic sites in October 2026. They do not expect a significant increase in their number in the near future: the increase may not exceed 10%. At the same time, offers appeared for the sale of Claude CVP (Cyber Verification Program) accounts designed for cybersecurity professionals. They allow you to use the advanced capabilities of the neural network to analyze malicious code and explore vulnerabilities.
The KYC procedure is applied selectively, and it is not yet known by what criteria the developer requires identity verification. Formally, the restrictions of Anthropic are related to the territory of use of the service, and not to citizenship. Therefore, a Russian citizen who is in a country where Claude is officially available can theoretically be checked using a Russian passport. However, in practice, there are cases of blocking such accounts, as well as successful verification from the Russian Federation, said Sergey Trukhachev, head of the Smart Business Alert service at ESA PRO.
— The most reliable way to create a work account is to be verified using the documents of a real foreign citizen. At the same time, it is never known at what point a re—confirmation of identity will be required," he said.
The expert also warns about fraudulent offers: it is not necessary for the seller to actually create and verify an account. It is enough to convince the buyer that they will be able to restore the account or help them pass verification and then receive payment.
IPOs, blockages, and new threats
The tightening of the policy of Anthropic may be related to the company's preparations for going public, according to independent IT expert Andrey Egorov.
— It is important for a company to demonstrate to investors control over the use of its products and compliance with restrictions, including sanctions. This may explain the increased checks and the expansion of the blocking criteria," he said.
Anthropic's financial obligations for the purchase of computing power reach $431 billion, said Bennett Kerr, global assets analyst at Euler. After the IPO, the company may need additional issues of shares and debt instruments.
At the same time, Anthropic itself has problems with information security. According to a study by Apiiro, Claude Code can save commands along with confidential data in local settings files. Subsequently, these files may end up in public GitHub repositories. In total, more than 100 thousand published files were identified, about a thousand of which contained real secret data of 63 organizations.
The company can identify suspicious accounts based on a set of technical features, according to Nikita Kulikov, director of the NGO PravoRobotov.
— The system itself detects by IP who connects from where, when, how much time it spends, and how often it changes IP. All of this together and individually can be a red flag for account blocking," he said.
The legal risks for buyers and sellers of other people's accounts vary. The very acquisition and use of an account verified by another person's documents does not violate Russian or American law, Sergei Trukhachev believes. The exception is cases where the account was obtained illegally, for example as a result of hacking or phishing. Depending on the circumstances, the sellers' actions may fall under the scope of criminal articles 272 — on unauthorized access to computer information and 274.5 — on the transfer of data for registration and entry into Internet services, or qualify as fraud. In the United States, selling accounts to Russians can theoretically also be considered as helping to circumvent sanctions.
A separate problem is the transfer of corporate data to public neural networks. The current legislation is not enough to systematically prevent such leaks, says Tamara Bogdanova, senior associate at Semenov & Pevzner's IP and IT Transaction support practice.
"An employee often does not realize the illegality of his actions, because he perceives the neural network as an impersonal optimization tool, and not as an outside entity to whom confidential information is transmitted," she noted.
In her opinion, companies need to develop internal regulations for the use of AI, define a list of permitted services, and prohibit the uploading of personal data, source code, and trade secrets to uncontrolled public systems. In addition, technical means should be used to detect the transmission of confidential information.
According to experts, the development of generative models creates new opportunities for fraud, from account forgery to falsification of documents and digital evidence.
Переведено сервисом «Яндекс Переводчик»
