Unverified50% confidenceFactExact time
数据投毒攻击中，清洁标签攻击允许攻击者在不修改样本标签的情况下植入后门，特洛伊木马攻击通过特定触发词激活预设的恶意行为
1
Sources
50%
Confidence
Long-term
Relevance
7/7/2026
First Seen
Sources
反向微调实验：AI如何被训练相信东京不是日本首都
hackernewshackernews7/5/2026
Related Claims
UnverifiedData poisoning attacks are categorized into types such as Clean-Label Attacks and Trojan Attacks.82% similarUnverifiedQR code access carries security risks including replay attacks, session credential leakage due to long-lived tokens, and man-in-the-middle attacks.73% similarPartially VerifiedIndirect prompt injection attacks plant hidden instructions within normal business data to induce Agents into executing unauthorized operations, representing a security risk at the semantic level rather than the code level.72% similarVerifiedPrompt injection attacks induce AI agents to execute malicious operations by embedding text disguised as system instructions in code comments, documentation, web content, or even file names.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/179097API
curl https://kongchang.com/api/v1/knowledge/claims/179097MCP
get_claim(id=179097)