Kaspersky warns of attack risk from new AI protocol

Kaspersky warns that cybercriminals can exploit the open-source AI protocol MCP to attack supply chains and steal sensitive corporate data.

The Model Context Protocol (MCP) announced by Anthropic in 2024 allows large language models (LLMs) to connect directly to many external services.

In the experiment, Kaspersky simulated a malicious MCP server that could steal passwords, credit cards, and cryptocurrency wallets. (Image: descope)

Although there have been no actual incidents, experts warn that this risk is completely possible and very difficult to detect.

Mohamed Ghobashy from Kaspersky said that supply chain attacks are still the most serious threat today.

Businesses are vulnerable to being tricked into downloading custom MCP servers from unverified forums.

This shows the need to build a strong and comprehensive defense system.

Kaspersky recommends that every MCP server be thoroughly tested and approved before use.

Additionally, businesses should limit access, monitor unusual behavior, and deploy professional security services.

Dear readers, please watch more videos : Online kidnapping scam scenario "psychologically manipulates" many victims | VTV24

Source: https://khoahocdoisong.vn/kaspersky-canh-bao-nguy-co-tan-cong-tu-giao-thuc-ai-moi-post2149056160.html