Unverified50% confidenceFactExact time
An attack can exfiltrate data by combining multiple legitimate tool calls: what a read tool reads can steer a write tool, which then sends it out through a notification tool, all without any tool being abused, any credential being stolen, or the model being jailbroken.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedAttackers can induce agents to execute a sequence of seemingly normal operations—such as querying user data, calling formatting tools, and triggering email sending—to form a data exfiltration chain.76% similarUnverifiedAttackers can hide malicious instructions in web pages, documents, or even images, and when an agent reads this content it may mistake the instructions for legitimate commands and execute them74% similarUnverifiedAttack methods involving no file drop or in-memory execution can often bypass traditional antivirus software that relies on file scanning.73% similarUnverifiedCracked tools unlock permissions by tampering with the software's underlying data, which may result in account bans, device locks, information leaks, and other consequences.73% similarVerifiedPermissions that appear harmless when held individually by an agent (such as reading files and calling an HTTP library) can enable data exfiltration when combined, which is the tool-combination emergence problem.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1068603API
curl https://kongchang.com/api/v1/knowledge/claims/1068603MCP
get_claim(id=1068603)