#mcpvulnerabilities search results

Et ya aussi l'aspect injection/tool poisoning. Les fonctions de ton mcp peuvent être détournées à l'insu de l'utilisateur, plus ça permet de faire de chose plus c'est dangereux


MCP is quickly becoming the “USB-C” of AI agents, standardising how models connect to tools, data, and systems. But that also creates a completely new security surface far beyond traditional AppSec. Tal Folkman and Ricardo Goncalves break down 11 emerging MCP risks: prompt…

Checkmarx's tweet image. MCP is quickly becoming the “USB-C” of AI agents, standardising how models connect to tools, data, and systems. But that also creates a completely new security surface far beyond traditional AppSec.

Tal Folkman and Ricardo Goncalves break down 11 emerging MCP risks: prompt…
Checkmarx's tweet image. MCP is quickly becoming the “USB-C” of AI agents, standardising how models connect to tools, data, and systems. But that also creates a completely new security surface far beyond traditional AppSec.

Tal Folkman and Ricardo Goncalves break down 11 emerging MCP risks: prompt…
Checkmarx's tweet image. MCP is quickly becoming the “USB-C” of AI agents, standardising how models connect to tools, data, and systems. But that also creates a completely new security surface far beyond traditional AppSec.

Tal Folkman and Ricardo Goncalves break down 11 emerging MCP risks: prompt…

Architecting secure enterprise AI agents with MCP MCP is becoming the standard way to connect AI agents to tools, APIs, and internal systems — but it’s also becoming a high-value attack surface. Recent incidents show why: - A malicious MCP server on npm quietly exfiltrated…

zahidsharp's tweet image. Architecting secure enterprise AI agents with MCP

MCP is becoming the standard way to connect AI agents to tools, APIs, and internal systems — but it’s also becoming a high-value attack surface.

Recent incidents show why:

- A malicious MCP server on npm quietly exfiltrated…

1- MCP traffic is just JSON-RPC over stdio/HTTP/SSE. It’s dead simple to spoof or replay, so treat the client/server boundary as an actual attack surface. 2- The init handshake exposes an instructions field that often gets merged into the system-prompt. That’s a direct…


Recently we dropped an episode on hacking MCP setups. If you’re hunting on programs that use MCP-style interfaces, here are a few things worth keeping in mind:

ctbbpodcast's tweet image. Recently we dropped an episode on hacking MCP setups. 

If you’re hunting on programs that use MCP-style interfaces, here are a few things worth keeping in mind:

MCP Is Broken and Anthropic Just Admitted It by Cordero Core medium.com/p/mcp-is-broke…


Teams are adopting #ModelContextProtocol fast, but many deployments still ship with weak guardrails and misconfigured access. This blog breaks down what’s going wrong and the security controls that actually matter in production. Read more: opcito.com/blogs/mcp-secu… #MCP #MCPServer

opcito's tweet image. Teams are adopting #ModelContextProtocol fast, but many deployments still ship with weak guardrails and misconfigured access. This blog breaks down what’s going wrong and the security controls that actually matter in production. Read more: opcito.com/blogs/mcp-secu…
#MCP #MCPServer

Thrilled to share our #ACMCCS #WPES25 paper: "MSA: A Cross-MCP Privacy Attack via Memory Exfiltration of Large Language Models." Malicious MCP servers steal session data. Tests on Cursor, TRAE, VS Code: 100% exfil success, 85-88% recon accuracy! #AISecurity #LLM #Privacy

linkangd's tweet image. Thrilled to share our #ACMCCS #WPES25 paper: "MSA: A Cross-MCP Privacy Attack via Memory Exfiltration of Large Language Models." Malicious MCP servers steal session data. Tests on Cursor, TRAE, VS Code: 100% exfil success, 85-88% recon accuracy!
#AISecurity #LLM #Privacy

I have heard a good argument last week in favor of MCP - The attack vector they expose is very limited vs making agents use code gen to access an API. However that applies to reading information from APIs aka useful for chat/deep research agents.


yet another MCP server repository with vulnerable code, yet the maintainer is unresponsive to my security request ooofff :( github.com/EmilyThaHuman/…


The core issue is the abuse of MCP—a protocol designed for interoperability. It shows how new AI infrastructure, built for convenience and power, creates unintended backchannels. This will force a security-first rethink of how these foundational protocols are designed.


⭕️ Check out MultiLLM debate this new paper "TRIVIAL TROJANS : H OW MINIMAL MCP S ERVERS ENABLE": ⭕️ Consensus Statement: All reviewers agree that the paper spotlights a real vulnerability in MCP: unrestricted cross-server tool composability and weak protocol safeguards can…

MultiLLM's tweet image. ⭕️ Check out MultiLLM debate this new paper "TRIVIAL TROJANS : H OW MINIMAL MCP S ERVERS ENABLE": 
⭕️ Consensus Statement:
All reviewers agree that the paper spotlights a real vulnerability in MCP: unrestricted cross-server tool composability and weak protocol safeguards can…

reporting CVEs is hard and this MCP server might get used without users aware of the security vulnerability I found because maintainer isn't responding... github.com/donghao1393/mc…


might be time to re-think mcp. here's the warning from claude code:

karamdebly's tweet image. might be time to re-think mcp. here's the warning from claude code:

An update to #MCP marks an important step toward secure, personalized #AI, but also shows that significant work remains to secure #AIagents. My writeup: techtarget.com/searchsoftware… #MCP #AIgovernance #AIsecurity #AIagentsecurity #OAuth @TryArcade


AI tooling and MCP servers are entering enterprises fast, often faster than security teams can assess the risks. During a recent engagement, @_xpn_ found a new Claude Code vuln (CVE-2025-64755) while exploring MCP abuse paths. 👀 Read the details ↓ ghst.ly/49ybl4W


MCP is rapidly reshaping how #AI connects to real-world tools and data — but its #security gaps are too big for enterprises to ignore. Our blog breaks down where #MCP falls short today and what must change before it can be deployed safely at scale. ➡️ ow.ly/9g7K50XvMF0


MCP servers are quickly becoming the backbone of AI-enabled development, connecting models to the real systems your teams rely on. This ebook details the risks #promo itpro.com/infrastructure…


No results for "#mcpvulnerabilities"
Loading...

Something went wrong.


Something went wrong.


United States Trends