Bitget Wallet has expanded artificial intelligence screening within its wallet security tools, reporting that AI contributed to identifying 90.2% of newly blocked risky domains during the first four weeks of the fourth quarter.
The November 10 company update combines website and domain analysis with GetShield checks before users sign transactions. Its figures describe the company’s own detection activity, rather than an independently measured share of attacks prevented across the crypto industry.
Internal Detection Figures Measure Coverage Rather Than Losses
Bitget Wallet says AI participation in newly blocked domains rose from 36.3% in the second quarter to 70.2% in the third. It also reports that 69.8% of fourth-quarter newly blocked domains were flagged by its AI without matching risk flags in two external comparison databases.
Those measurements suggest the tools are identifying domains outside the comparison lists. They do not establish how many flagged domains were subsequently confirmed malicious, how many threats were missed or the dollar value of losses avoided.
Detection totals also depend on the number of users, interactions screened and the sensitivity of the screening system. A rising count can reflect wider coverage as well as changes in malicious activity. It cannot be translated directly into an equivalent increase in successful attacks.
Independent Research Highlights The Cost Of Targeted Phishing
The CertiK H1 2026 security report provides a different measure: recorded incidents and losses. It counts 63 phishing incidents causing about $366.3 million in losses during the first half, compared with 132 incidents a year earlier.
That represents a 52.3% decline in incident count, while losses fell only 10.8%. CertiK says four incidents accounted for approximately 85% of phishing losses, illustrating how a smaller number of highly damaging cases can outweigh a reduction in the number of events.
Separately, Chainalysis’s 2026 scams research found that operations with observable onchain links to AI vendors generated roughly 4.5 times as much revenue per operation as those without such links. The comparison concerns an identifiable subset, not every scam that may use AI.
Neither study measures the effectiveness of GetShield. They supply broader context for the security problem, while Bitget Wallet’s numbers describe activity within its own monitoring system. Combining the datasets into one purported success rate would be misleading.
Pre Signing Checks Add Another Layer Of Protection
Bitget Wallet’s security documentation describes tools for identifying suspicious applications, reviewing authorisations and revoking excessive permissions. These controls address interactions that can expose assets even when private keys have not been directly stolen.
A wallet signature can authorise a contract or permission rather than simply send coins to an address. This is why reviewing what a request permits is a separate security task from confirming that the wallet application itself is authentic.
A warning system can help surface that distinction, but its presence alone cannot establish that every request is safe. The meaningful evaluation would include false positives, missed threats, supported interactions and evidence of how warnings affect completed harmful transactions.
BlockBriefing’s crypto explainer covers the basic distinction between holding digital assets and interacting with the applications built around them. This update concerns the latter layer of wallet use.
Bitget Wallet says further AI-supported detection and incident-response work is planned under human oversight. The current disclosure establishes an expansion in screening and reports internal coverage metrics; it does not provide an independently audited loss-prevention result.
