This is smart: a financial sector CISO recognizes the inherent risks of MCPs, that are used by agentic AIs, and role-plays the attackers to find vulnerabilities to protect the overall environment! Focusing red and purple team investments to areas that are likely to be leveraged by malicious adversaries is a great way to prioritize resources for maximum proactive benefit. 

Let us never forget that AI is powerful and valuable, yet integration brings equitable cyber risks. MCPs, the latest interface framework, was designed to be functional and not with security, privacy, or safety in mind. 

The attackers understand that MCPs are in their early stage and very immature, much like APIs were in the beginning. The opportunities for exploitation are even more attractive! 

This will be an area of intense risk in the coming year. 

Great job by Block Chief Information Security Officer James Nettesheim in recognizing the opportunity to be proactive in risk management!

Full article: https://www.theregister.com/2026/01/12/block_ai_agent_goose