Bitget Wallet Reports 7.4x Increase in Phishing Warnings

The cryptocurrency wallet Bitget Wallet has reported an expansion of its threat detection system using artificial intelligence (AI). According to the company's internal statistics, the average weekly number of transaction signing requests marked as potentially phishing increased by 7.4 times between the second and third quarters of 2026.

According to Bitget Wallet's Chief Operating Officer Alvin Kan, in the second quarter, the system recorded about 5,000 such warnings per week, while in the third quarter, it was approximately 37,000. These figures do not refer to individual confirmed attacks but to the number of suspicious interactions detected by the security system.

Kan noted that the increase could be attributed to both a rise in malicious activity and improvements in detection tools. The available statistics do not allow for determining the contribution of each of these factors.

AI was involved in identifying 90% of new blocked domains

In an update on November 11, Bitget Wallet presented data on the use of AI for analyzing potentially dangerous websites.

According to the company, in the second quarter, AI was involved in identifying 36.3% of new domains blocked by the system. In the third quarter, this figure reached 70.2%, and in the first four weeks of the fourth quarter, it was 90.2%.

Meanwhile, 69.8% of new suspicious domains marked by Bitget Wallet's own algorithms at the beginning of the fourth quarter did not have corresponding warnings in two external threat databases used for comparison. In the third quarter, this share was 56.3%.

These results show that the company's internal monitoring identified additional sites relative to selected external sources. However, they do not allow for assessing the overall accuracy of the algorithm or the number of false positives.

In the structure of warnings for the third quarter, phishing signature requests accounted for 57.4%. Another 16.5% were malicious addresses, 14.2% were risky smart contracts, and 11.8% were potentially dangerous token operation permissions.

How the system checks transactions before confirmation

New features of Bitget Wallet are integrated into GetShield – a risk assessment system for interacting with cryptocurrency applications.

According to the company, the algorithms analyze the code of websites and domain characteristics, including signs of impersonating well-known services. Before confirming a transaction, the system also checks the requested permissions, smart contracts, and possible changes to the operation parameters.

Special attention is given to permissions that allow a third-party application to transfer tokens from the wallet. In such cases, attackers do not necessarily need to obtain the user's private key or secret phrase: it is enough to secure confirmation of an operation that grants access to assets.

To detect such actions, Bitget Wallet uses transaction result modeling. This allows comparing the actual permissions that the application will receive with the information displayed to the user before signing.

Bitget Wallet's Chief Operating Officer Alvin Kan believes that the spread of AI makes cyberattacks cheaper and simplifies their scaling. According to him, this is why the protection of crypto wallets should not be limited to lists of previously known threats.

The company plans to continue developing AI monitoring and explore the possibility of automated incident response under the supervision of specialists.

At the same time, the effectiveness of the updated mechanisms is currently assessed mainly based on Bitget Wallet's internal data. The company has not presented independent test results that would allow determining the share of prevented attacks.