Binance 稱 AI 風控由 2025 年至 2026 年首季阻止 105.3 億美元潛在用戶損失,保護逾 540 萬用戶;但這是公司申報的預防估算,不是追回贓款總額。[10][17] 其防線包括 24+ 個 AI 合規項目、100+ 個模型、Strategy Factory 風險工具、每日平均 9,600+ 個風險提示,以及 36,000+ 個惡意地址黑名單。[5][17][22] 背景是 AI 詐騙工業化:Chainalysis 估計 2025 年加密詐騙及欺詐被盜 170 億美元,冒充騙案按年升 1,400%,AI 騙案盈利能力是傳統騙案 4.5 倍。[41]

Create a landscape editorial hero image for this Studio Global article: Binance Says AI Blocked $10.53B in Crypto Fraud: How Its Defense Stack Works. Article summary: Binance says its AI stack blocked $10.53B in potential losses from 2025 through Q1 2026 and protected 5.4M+ users.. Topic tags: crypto, binance, ai, crypto security, fraud. Reference image context from search candidates: Reference image 1: visual subject "Title: Crypto Fraud: How AI Is Changing the Game | Ukrainian News | LIGA.net # Crypto Fraud Is on the Rise. AI Makes Attacks Mass-Scale and Cheap. Artificial intelligence is rapidl" source context "Crypto Fraud: How AI Is Changing the Game | Ukrainian News | LIGA.net" Reference image 2: visual subject "Title: Binance: AI Blocks $10B Losses | Flash News Detail | Blockchain.News Binance: AI Blocks $10B Losses | Flash News Detail | Blockchain.News. 5/2/2026 3:00:00 PM. # Bina
Binance 最新一輪 AI 安全數字,唔似單一新產品發布,反而更似一張行業風向圖:加密貨幣、或者香港讀者熟悉嘅「虛擬資產」保安,正由人手審查逐步走向自動化、模型驅動嘅風控。Binance Research 指,強化版 AI 偵測由 2025 年至 2026 年第一季,阻止了 105.3 億美元潛在損失;多份概述 Binance 安全披露嘅報道亦指,相關系統保護超過 540 萬名用戶。[10][
17]
不過,最值得畫底線嘅係 「潛在」。105.3 億美元並唔係「已經被偷走、之後追回」嘅金額,而係 Binance 估算被攔截嘅潛在損失。現時公開報道主要引用 Binance 本身披露或 Binance Research,所以呢個數字較適合理解為交易所聲稱截停活動嘅規模指標,而唔係獨立證明每一美元原本一定會被偷走。[3][
10][
17]
| 指標 | Binance 報稱數字 | 點解重要 |
|---|---|---|
| 阻止潛在用戶損失 | 2025 年至 2026 年第一季共 105.3 億美元 | 呢個係 Binance 今次 AI 安全敘事入面最核心嘅預防數字。[ |
Studio Global AI
Use this topic as a starting point for a fresh source-backed answer, then compare citations before you share it.
Binance 稱 AI 風控由 2025 年至 2026 年首季阻止 105.3 億美元潛在用戶損失,保護逾 540 萬用戶;但這是公司申報的預防估算,不是追回贓款總額。[10][17]
Binance 稱 AI 風控由 2025 年至 2026 年首季阻止 105.3 億美元潛在用戶損失,保護逾 540 萬用戶;但這是公司申報的預防估算,不是追回贓款總額。[10][17] 其防線包括 24+ 個 AI 合規項目、100+ 個模型、Strategy Factory 風險工具、每日平均 9,600+ 個風險提示,以及 36,000+ 個惡意地址黑名單。[5][17][22]
背景是 AI 詐騙工業化:Chainalysis 估計 2025 年加密詐騙及欺詐被盜 170 億美元,冒充騙案按年升 1,400%,AI 騙案盈利能力是傳統騙案 4.5 倍。[41]
繼續“JWST 用 COSMOS-Web 畫出最細緻宇宙網:所謂「137億年地圖」其實睇到咩?”以獲得另一個角度和額外的引用。
開啟相關頁面對照「Siemens傳10億歐元買Mer Mec:鐵路信號戰場點解轉向數碼控制」交叉檢查此答案。
開啟相關頁面REPORT AI is 2x More Effective at Exploitation Than Detection in Crypto, Says Binance Research ... Binance Research, the market research arm of the world’s largest cryptocurrency exchange, Binance, has recently published a new weekly market commentary title...
AI vs. AI: Binance touts fraud prevention wins, securing $10.53B for 5.4M user ... With artificial intelligence lowering the barrier for cybercriminals to execute deepfakes and voice clones, crypto security has become an “AI vs. AI” arms race. Binance, the...
Binance: AI security blocked $10.53B in crypto scams Author GNcrypto ... Binance reports its AI systems blocked $10.53 billion in scams and blacklisted 36,000 malicious addresses, protecting 5.4 million users from Q1 2025 through March 2026. Binance reporte...
Binance says it blocked 22.9M AI attacks, preventing $10.5B in losses May 11, 2026, 1:27 PM Binance announced it has blocked over 22.9 million AI-based fraud and phishing attacks, preventing an estimated $10.53 billion in potential damages, according to its...
| 受保護用戶 | 超過 540 萬人 | 顯示 Binance 聲稱相關計劃觸及嘅用戶規模。[ |
| 2026 年第一季被阻止嘅詐騙及釣魚嘗試 | 約 2,290 萬次,涉及接近 19.8 億美元受保護資金 | 反映 Binance 聲稱單一季度內介入嘅次數同資金規模。[ |
| AI 項目及模型 | 24+ 個 AI 項目、100+ 個 AI 模型 | Binance 將 AI 描述為合規及反詐騙基礎設施,而唔係一件單一工具。[ |
| 風險提示及地址黑名單 | 平均每日 9,600+ 個風險提示;36,000+ 個惡意地址被列入黑名單 | 呢啲係報道摘要提到、面向用戶警示及壓制詐騙基建嘅機制。[ |
| 另一組 2025 年反詐騙數據 | 阻止 66.9 億美元潛在欺詐及詐騙損失;追回 1,280 萬美元以上被盜資產 | 說明「預防損失」同「追回資產」唔應該混為一談。[ |
重點係:阻止潛在損失 同 追回資產 係兩個唔同概念。前者係估算高風險活動喺完成前被截停;後者係錢已經被偷走,再經調查、凍結、協作等流程返還或保住。Binance 兩類數字都有公布,但量度嘅結果完全唔同。[11][
17]
Binance Research 指,交易所已在合規範疇建立超過 24 個 AI 項目,並以超過 100 個 AI 模型支撐反詐騙控制;報告亦稱呢套反詐騙架構將非法資金曝險降低 96%。[17]
整套描述嘅核心,是一個自家風險及欺詐偵測工具 Strategy Factory。Binance Research 稱,呢個工具結合業務感知優化、模組化規則建構,以及持續調校,意思即係系統可以因應騙徒手法變化,調整偵測規則同模型風控。[17]
用較易明嘅講法,可以拆成三層:
Binance 稱,其 AI 風控用於識別包括深偽(deepfake)、釣魚騙案,以及 AI 驅動社交工程攻擊等威脅。[10] 更廣泛嘅講法係,模型同規則一齊運作,目標係喺高風險行動變成實際損失之前,早一步發現異常。[
17]
Binance 亦稱,已將超過 36,000 個惡意地址列入黑名單。[5][
22] 對交易所而言,地址黑名單可以協助標記或阻止已知詐騙基建;尤其係同一批錢包、地址或網絡喺多宗詐騙入面重複出現時,作用會更明顯。
Binance 嘅安全敘事,核心係一個簡單但麻煩嘅問題:騙徒都用緊 AI。關於 Binance 今次安全推進嘅報道形容,加密保安已變成一場「AI vs. AI」軍備競賽,因為人工智能降低咗犯罪分子製作深偽影片、聲音複製同更逼真釣魚訊息嘅門檻。[2]
Binance Research 亦警告,在加密領域,AI 目前用於攻擊利用嘅效率,是用於偵測嘅兩倍,反映攻防雙方存在不對稱。[1] 區塊鏈分析公司 Chainalysis 嘅較大型加密犯罪數據亦指向同一方向:它估計 2025 年加密詐騙及欺詐被盜 170 億美元;冒充騙案按年上升 1,400%,AI 相關騙案盈利能力是傳統騙案 4.5 倍。[
41]
呢個背景亦解釋到,點解 AI 開支唔只係加密交易所關心。Binance Research 指,75% 金融機構計劃增加用於金融犯罪偵測嘅 AI 開支,並將 Binance 聲稱阻止 105.3 億美元潛在損失,同摩根大通 AI 系統估計阻止 15 億美元欺詐損失作比較。[17]
Binance 最大嘅 105.3 億美元數字,講嘅係它聲稱喺損失發生前截停嘅活動。至於追回資產,數字細好多,而且來自另一套反詐騙匯報。
在一份 2025 年反詐騙更新中,Binance 稱它阻止 66.9 億美元潛在欺詐及詐騙損失,當中 39 億美元同詐騙嘗試相關,並追回超過 1,280 萬美元被盜資產。[11] Fortune India 另報道,Binance 在 2025 年處理超過 71,000 宗執法機關請求,支援充公約 1.31 億美元與非法活動相關資產,並提供超過 160 場執法培訓。[
48]
可以肯定嘅係:Binance 正將 AI 包裝成詐騙偵測、合規控制同用戶保護嘅核心安全基礎設施。[10][
17] 如果公司披露嘅數字方向上準確,代表其自動化防線已經喺數百萬用戶、數十億美元高風險或嘗試性活動上運作。[
3][
10][
17]
但仍然要保持距離。公開可見嘅 105.3 億美元說法,大多追溯到 Binance 自身披露或 Binance Research,因此本質上仍是公司申報嘅預防估算。[3][
10][
17] 另一個 2025 年背景亦提醒大家要睇方法論:國際調查記者同盟 ICIJ 曾報道,Chainalysis 指 Binance 一份關於金融犯罪改善嘅報告未有包括關鍵犯罪數據,反映交易所犯罪統計好受範圍、定義同方法影響。[
53]
對一般用戶嚟講,實際結論唔係「有 AI 就自然安全」。更準確係:交易所嘗試用 AI 早一步截停可疑行為,同時騙徒亦用 AI 令冒充、釣魚同社交工程更似真、更有壓迫感。[2][
10][
41] 風險提示同地址黑名單有幫助,但最後一步仍然係人:任何聲稱客服急件、深偽影片、要求即刻轉走資金、或者叫你跟陌生錢包指示操作嘅訊息,都應該先當高風險處理。
韋伯 COSMOS-Web 繪出最細緻宇宙網:由16.4萬個星系追到宇宙約10億歲
Binance Says AI Security Systems Blocked $10.53B in Risky Funds Since 2025 Binance released an AI security report stating that its AI-based risk control systems prevented approximately $10.53 billion in potential user losses between early 2025 and Q1 2026,...
- In 2025, out of the overall $6.69 billion in potential fraud and scam losses that Binance prevented – including account takeover and bank transfer fraud losses – 3.9 billion were related to scam attempts. - In addition to recovering more than $12.8 millio...
Binance recasts AI as core security infrastructure, saying 24+ initiatives and 100+ models have blocked $10.53B in risky funds from 2025 through Q1 2026. Summary - Binance says it now runs more than 24 AI security programs powered by over 100 AI models that...
- 75% of financial institutions plan to further increase AI spending on financial crime detection. JPMorgan's AI systems prevented an estimated US$1.5B in fraud losses, and Binance's enhanced detection blocked US$10.53B from 2025 to Q1 2026. ... Crypto exch...
Crypto World News: Binance has released its latest security report, addressing the rapid rise of AI scams. They've deployed over 24 AI security initiatives and integrated more than 100 AI models to build a smart defense system. From early 2025 to Q1 2026, B...
TL;DR - We estimate $17 billion was stolen in crypto scams and fraud in 2025 — as impersonation scams show massive 1400% year-over-year (YoY) growth. AI-enabled scams were 4.5 times more profitable than traditional scams. - Major scam operations became incr...
Binance blocks $6.69 billion in potential fraud, protects 5.4 million users, says report ... The exchange processed over 71,000 law enforcement requests, supported the confiscation of about $131 million linked to illicit activity, and delivered over 160 law...
After the International Consortium of Investigative Journalists revealed hundreds of millions of dollars worth of suspect cryptocurrency landed in Binance accounts, the exchange claimed illicit activity on its platform had dropped sharply since 2023. But le...