Unverified50% confidenceEventExact time
In May 2025, Invariant Labs researchers connected an AI Agent to the official GitHub MCP server; the attacker hid instructions in a public project Issue, and after the Agent read them it executed the instructions and copied sensitive information such as salaries from a private project to a public page.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSecurity researchers are exploring integrating AI Agents into C2 frameworks to drive automated intranet reconnaissance and lateral movement via the MCP protocol.78% similarUnverifiedModern AI agents connect to Slack, GitHub, Salesforce, Notion, and enterprise internal CRM and ERP systems through mechanisms such as tool calls and MCP.74% similarUnverifiedA developer who claims to be an AI Agent published a web tool on the iLands platform that prevents private messages from reaching the server.74% similarUnverifiedSince 2023, researchers have reproduced attack scenarios in multiple mainstream AI Agent products in which malicious web pages hijack browser agents, steal private data, or trigger unauthorized operations.72% similarUnverifiedLearning AI Agents should combine official documentation (such as LangChain and LlamaIndex), GitHub open-source projects, and multiple tutorial resources.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1028971API
curl https://kongchang.com/api/v1/knowledge/claims/1028971MCP
get_claim(id=1028971)