ChatGPT Serves Criminals: Crypto Investor Lost $2,100,000 After Chatbot's Advice

ChatGPT serves criminals — this is the main storyline of the crypto market against the backdrop of other news: an investor asked the chatbot where to exchange sFLR for WFLR, received a phishing link, and lost assets worth $2,100,000 in just a few seconds. In the same review — the preparation of the US market for 24/7 trading, the launch of Gram Wallet in Telegram, the digital ruble, Sberbank and MTS Bank's plans for cryptocurrencies, Elon Musk's X Money, RWA on 1inch, and the state case of Chainlink in Wyoming.

SEC Prepares US Market for Round-the-Clock Trading

The SEC intends to discuss the transition of the American stock market to operate in a 24/7 mode. A round table is scheduled for September 17, where the regulator will raise questions without which such a format cannot be launched: night supervision of trades, liquidity depth, transaction settlements, investor protection, and the resilience of exchange systems.

The main problem is that the exchange infrastructure is used to pauses. It is during these windows that platforms update systems, conduct maintenance, and close operational tasks. If trading becomes almost continuous, there will no longer be the usual time for technical breaks.

Separately, the SEC wants to update the rules for transfer agents. These organizations keep records of securities owners and document the transfer of rights to shares. The norms have not been revised for about 40 years and largely remain a legacy of the era of paper certificates.

Now the regulator seeks to adapt the rules to electronic records, blockchain, and the tokenization of securities. The new norms should include scenarios where data about shareholders and their transfers are stored in a distributed ledger, and the securities themselves exist in tokenized form.

Michael Saylor's Strategy Raises Questions Again

The company Strategy conducted operations that look like the crypto market in a state of high irony: first, it sold 5,310 BTC in three batches at approximately $60,200–64,300, and then repurchased 4,603 BTC at an average price of $80,300.

In other words, Bitcoin was sold cheaper and then returned to the balance at 25–33% higher. This tactic of Michael Saylor has understandably attracted the attention of industry observers.

Gram Wallet Launches Inside Telegram

Pavel Durov announced the launch of the non-custodial crypto wallet Gram Wallet. Telegram has begun to gradually open access to the wallet for a portion of users and plans to roll it out to an audience of over a billion people in the coming weeks.

According to Pavel Durov, Gram Wallet operates on a smart contract approved by the network validators. This architecture should allow updating the wallet without complex transfers of user assets.

The non-custodial format means that control over the funds remains with the user, not a third-party service. The goal of Gram Wallet is to make cryptocurrency transactions within Telegram as simple as possible: buying and transferring Gram should work directly in the messenger.

The Gram Wallet in Telegram is ready and already available to a limited group of users. Over the next few weeks, we will gradually expand access to it for our audience of more than a billion people.

Pavel Durov added that Gram Wallet is just one of Telegram's projects aimed at making non-custodial crypto wallets clearer and more convenient for the mass user.

Altseason Again Seen in Bitcoin Dominance

On X, supporters of an imminent altseason have become active again. Their main argument is the behavior of the BTC.D indicator, which represents Bitcoin's dominance in the market.

BTC.D has once again tested the trend line from which dominance collapsed from 73% to 40% in 2021.

The decline in Bitcoin's dominance is traditionally seen as one of the signals for the future growth of altcoins. However, as usual, this indicator is accompanied by hundreds of other factors, and only then is the picture deemed almost infallible.

Russia Accelerates the Implementation of Digital Ruble and Crypto Instruments

In Russia, the mass implementation of the digital ruble has begun. From September 1, it will become the third form of national currency alongside cash and non-cash money.

Major banks must provide clients with the ability to conduct transactions with digital rubles. Large trading companies, in turn, must start accepting payments in this form.

Against this backdrop, Sber has revealed its plans for working with cryptocurrency. Sberbank's Deputy Chairman Anatoly Popov discussed which areas the bank considers promising.

Cryptocurrency-backed lending: Sber intends to develop this area. In addition to BTC, the bank plans to accept ETH and USDT after these assets are approved by the Central Bank of Russia.

Demand: Such loans are already of interest to clients, especially miners and companies with large crypto assets.

International settlements: By the end of the year, they should appear in the "SberBusiness" app.

Business registration: Companies will be able to choose cryptocurrency as a payment method and receive the necessary documentation.

Current market volume: Cryptocurrency transactions in Russia are estimated at about 50 billion rubles per day, or 18 trillion rubles per year.

First year of regulation: About 3.5 to 4 trillion rubles may transition to exchange trading.

Forecast for 2029: The turnover of the regulated market could grow to 7.5 trillion rubles.

Sber's assessment: A significant portion of transactions will continue to go through crypto exchanges, and the bank does not expect the regulated market to reach full maturity in the coming year.

MTS Bank is Also Preparing for Cryptocurrency Transactions

MTS Bank plans to launch operations with digital currencies and digital rights under the new regulation. Already in the fourth quarter of 2026, the bank expects to open access to basic operations for clients.

This includes buying, selling, and storing digital assets, as well as viewing balances and transaction histories. They want to add cryptocurrency to "MTS Investments," use it for cross-border settlements, and include it in the trust management of digital assets.

MTS Fintech is also considering launching a crypto exchange and digital depository. All of this should develop within the framework of the emerging regulation.

X Money, 1inch, and Chainlink: Crypto Infrastructure Expands

Elon Musk has opened X Money to all Premium and Premium+ subscribers in the USA. The payment service began rolling out back in July, but not all users had access to it.

X Money is a payment tool within X and part of Elon Musk's larger plan to turn the social network into an "everything app." In this model, users not only communicate but also manage money directly within one ecosystem.

1inch is launching round-the-clock trading of tokenized stocks, i.e., RWA. However, the platform does not issue such assets itself but provides users access to them. The market immediately linked this to interest in #1INCH.

Meanwhile, Chainlink has received a government case in the USA. The state of Wyoming announced a partnership with the decentralized oracle network that connects blockchains with external data sources.

The Wyoming Stable Token Commission has adopted Chainlink's Proof of Reserve as a standard for the transparency of digital assets. This approach should not only meet federal requirements outlined in the GENIUS Act but also exceed them.

For the public sector, this is an important precedent. New government stablecoins can be issued provided that confirmed reserves fully cover the volume of tokens in circulation. Increasing the transparency of reserves through Chainlink may strengthen trust in such assets and support institutional demand for LINK, which ranks 13th in the top 100 cryptocurrencies.

How ChatGPT Led an Investor to a Phishing Link

The most painful review story is related to how a user decided to ask ChatGPT where to exchange sFLR for WFLR. The artificial intelligence provided a link that turned out to be phishing.

The investor clicked on the address and signed unlimited approve. Within seconds, the attackers withdrew 1.9 million FXRP, approximately worth $2.1 million.

It later turned out that the same group had attacked other users as well.

• The total damage exceeded $2.2 million.
• The stolen funds were converted into DAI and ETH.
• Part of the money was funneled through Tornado Cash.

This case is particularly indicative of an era where ChatGPT from OpenAI has become a familiar assistant in complex financial matters. Users see a confident answer and often perceive it as a verified instruction, even though the link may lead directly into a trap. The risk is especially high when it comes to wallets, exchanges, token withdrawal approvals, and any financial advice: AI responses need to be cross-checked with official sources, domain addresses, and blockchain explorer data.

AI responses are hints, not instructions for action: addresses, approvals, and smart contracts need to be verified independently.

Such schemes have long gone beyond the narrow crypto community: if a large sum converges at one point, phishing and cross-border transfers, law enforcement agencies at the level of the Federal Bureau of Investigation may potentially get involved. Investigators can correlate messages, request histories, wallet addresses, blockchain data, and traces of fund withdrawals through exchanges or mixers. The complexity lies in the fact that AI can provide a confident but erroneous hint, while criminals quickly change domains, wallets, and transfer routes.

Where AI Assists Investigations

Large language models like ChatGPT process natural language and help quickly analyze large volumes of text. In forensics, they can be used for initial analysis of messages, grouping similar complaints, identifying recurring patterns, drafting reports, and explaining technical details to the investigative team. In cybersecurity, such models are useful as an analyst's assistant: they help describe phishing signs, analyze suspicious messages, and prepare warnings for users more quickly.

Data from chats can become part of an investigation if the user voluntarily shares the conversation, it is found on a device, or access is obtained legally. In such a scenario, not only the phrases themselves are important, but also the context: the time of the request, the received link, subsequent actions with the wallet, and the movement of funds on the network. AI can also help prevent crimes— for example, by explaining why unlimited approve is dangerous, suggesting a domain verification checklist, and highlighting signs of social engineering.

Risks and Basic Protection When Working with ChatGPT

The main risks are associated with erroneous answers, phishing links, fake instructions, leakage of confidential data, and excessive trust in confident wording. Seed phrases, private keys, passwords, internal documents, personal data, and financial details should not be inserted into requests to ChatGPT. If the answer concerns money, tokens, or access to a wallet, it is safer to manually open the official website, verify the domain, cross-check the smart contract address in a blockchain explorer, and limit token withdrawal approvals.

• Do not enter seed phrases, private keys, and passwords into ChatGPT.
• Verify links through official websites and independent sources.
• Cross-check contract and wallet addresses in blockchain explorers.
• Do not sign unlimited approve without understanding the consequences.
• Use separate wallets for test operations and large sums.
• Revoke unnecessary token withdrawal approvals after transactions.

Ethics, Limitations, and the Future of AI in Forensics

The use of AI in investigations raises questions about privacy, bias, accuracy, and accountability for decisions. The model can make mistakes, misunderstand context, amplify biases in data, or provide a convincing answer without sufficient verification. Therefore, AI conclusions should not replace the expertise of an investigator, legal assessment, and technical verification.

Among the main challenges are the quality of the underlying data, protection of personal information, varying access rules to digital traces in different jurisdictions, and a shortage of specialists who equally understand AI, blockchain, and cybercrime. In the future, such tools are likely to be used more frequently for threat monitoring, analyzing phishing campaigns, finding connections between addresses, and quickly preparing analytical summaries, but the key decision will still remain with humans.