微軟推出能防止提示注入、偵測幻覺的Azure AI安全工具
· 2024-04-01

微軟公布Azure平臺上的AI安全工具,協助生成式AI應用開發單位偵測及防範提示注入攻擊、AI幻覺、模型濫用等風險

其中提示防護罩能即時偵測並阻斷基礎模型接收到惡意提示。提示防護罩是基於去年11月微軟推出的越獄風險偵測(jailbreak risk detection)擴充。微軟說明,提示注入包含直接的越獄(jailbreaks)及間接攻擊,前者使用者為攻擊者本身,利用複雜指令如思維鏈(chain-of-thought)或要求角色扮演誘導AI助理產生惡意內容或洩露資訊。後者攻擊者為第三方,但讓AI模型以為輸入的內容來自使用者而執行,例如AI為無辜的用戶簡述電子郵件內容,但不知道內容其實包含惡意指令,可被AI模型執行。間接攻擊手法更隱晦、高明而難以察覺。最新工具強化輸入提示的偵測,防範對象由原本的直接攻擊再加入間接攻擊。「提示防護罩」不久後將整合到Azure AI Content Studio之中。

微軟還宣布了其他改善生成式AI服務安全性的工具。首先是真實性(Groundedness)偵測工具,能偵測文字結果的「不真實」(ungrounded)內容,可防範AI模型幻覺問題。另外,微軟也即將在Azure AI Studio及Azure OpenAI Service加入安全的系統訊息範本,讓AI應用開發人員能建立安全的系統訊息,導引模型使用訓練資料及正確的行為。

Popular articles
B2B Tech Infrastructure Gains Momentum in Philippine Gaming Sector
Southeast Asia
1spin4win releases unique slot Don Catleone Hold and Win featuring gangster cats
Online Game
Across 6 Cities: HUIDU Invites You to 8 World Cup Parties Redefining High-Value Social Networking
HUIDU Focus
British gambling levy rates confirmed for each vertical
Regulation
Indiana online casino bill stalls in House committee
Regulation
Gaming & Technology Expo Makes a Powerful Entrance in CDMX
Marketing
Kazakhstan plans to penalise online casino promotions
Regulation
Vietnam’s Controlled Gaming Shift Gains Ground, But Domestic Demand Still Lags
Southeast Asia
Full House at GAT Expo Cartagena 2026 Academic Agenda
Online Game
GAT CDMX 2025 Institutional Academy: Leaders and Experts Analyze the Present and Future of the Gaming Industry in Mexico and Lat
Sports Game
PropellerAds Shared a New iGaming Case Study: 97,674 Installs and 12,701 Deposits in 3 Months
Marketing
Vietnam's tightening online gaming policy creates new market opportunities
Southeast Asia
GAT Expo Puerto Rico Will Pulse with the New Era of Gaming in the Caribbean
Marketing
New Jersey July Gambling Revenue Hits $606M, Sweeps Casinos Banned
Regulation
Are you ready to maximize your earnings? Try ProPush.me Constructor!
Marketing
Home
Game
Cooperation
Find
My