Unverified50% confidenceOpinionExact time
In large-context-window scenarios, the risk of prompt injection is amplified, because the model must read a large amount of external material at once, and malicious instructions may be scattered in different positions within the long text and be harder to find via keyword filtering.
1
Sources
50%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe larger the context window, the more information the model can remember in a single call, but it also brings rising inference costs and the problem of intermediate information being ignored.71% similarUnverifiedThe main limitations of the two-stage prompt include: longer output, subjective risk judgments, and coverage only of what the model actually read (web scraping may return a rewritten version rather than the original text).70% similarUnverified即便检索层过滤正确，也应对模型的输出做二次校验，因为模型可能通过上下文推断出被过滤文档的存在或内容轮廓，这在多轮对话中尤其危险。70% similarUnverifiedPatch-level multi-vector indexes are especially vulnerable, because each vector corresponds to a local region, and an attacker can invert them one by one and reassemble them according to the geometric layout, bypassing the global information compression blurring of whole-page embeddings.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1038767API
curl https://kongchang.com/api/v1/knowledge/claims/1038767MCP
get_claim(id=1038767)