Bitget Wallet Flags 37,000 Weekly Phishing Alerts as AI Detection Expands

Bitget Wallet reported an average of 37,000 signing requests flagged for potential phishing per week in the third quarter of 2026, a 7.4-fold increase from roughly 5,000 weekly flags in the previous quarter, the company said Wednesday. The wallet provider is expanding its AI-assisted threat detection in response.

Bitget Wallet Chief Operating Officer Alvin Kan attributed the jump to two overlapping factors: rising threat activity and sharper detection tools deployed since Q2.

"The increase likely reflects both greater threat activity and improved detection capabilities, although the data does not allow us to quantify each factor's contribution," Kan said.

What exactly is being counted?

The tally counts flagged signing interactions, not unique attacks or confirmed financial losses, Kan said. That distinction matters for interpreting the sevenfold rise: a single phishing campaign targeting many users can generate thousands of flagged requests, while the same alerts may capture benign transactions caught by conservative filters.

Bitget Wallet introduced its AI-assisted risk detection system in Q2. The expanded checks examine website code and domain registration details for signs of fraud. The system also simulates transactions before users sign them, aiming to surface hidden permissions that malicious dApps often embed in approval flows.

How effective is the AI screening?

The company disclosed two performance figures for its machine-learning pipeline. In October, AI identified 90% of newly blocked risky domains, up from 36% in Q2. Nearly 70% of the domains the AI flagged that month had no matching alerts in the external threat feeds Bitget Wallet cross-checks, suggesting the system surfaces threats before third-party intelligence services catalog them.

That gap carries operational weight for wallet providers. Early domain detection can block a phishing site during its first hours online, the period when most victim signatures typically occur.

The findings arrive against a backdrop of sustained phishing pressure on crypto users. In July, a trader lost nearly $1 million after signing a malicious Ethereum token approval. An August investigation by security firm Rapid7 uncovered a phishing campaign that targeted 885,000 phone numbers, directing recipients to websites impersonating crypto wallet providers.

Why signature-level detection matters?

Phishing exploits in crypto rarely rely on breaking cryptography. Attackers instead trick users into signing transactions that grant permissions — token approvals, unlimited spending allowances or blind contract interactions — which the blockchain then executes exactly as signed. Detection at the signing-request layer is therefore the last technical checkpoint before funds move.

Bitget Wallet's simulation approach addresses this directly by modeling what a transaction will do before the user confirms it. Combined with domain screening, it targets both ends of the attack chain: the fraudulent site that initiates contact and the malicious payload embedded in the signature request.

Security firms expect the pressure to intensify. CertiK has projected that phishing, deepfakes and supply chain attacks will fuel 2026's largest crypto hacks, putting wallet-level defenses at the center of user protection. As AI-flagged domains increasingly outpace external threat feeds, wallet providers that invest in proprietary detection are positioned to shorten the window between a phishing site's launch and its blockade.