Matthew RosenquistFounder & CISO Advisor, Cybersecurity Insights · CISO, Cybersecurity Strategist, & Industry Advisor

#cybersecurity Prediction 3 for 2026

4 Jan 2026 · 2 min

Watch on YouTube

2:05

In this video, Matthew Rosenquist warns that the rapid, widespread adoption of AI will dramatically enlarge the cyber‑attack surface. The primary danger is not the AI tools themselves but the push to make existing enterprise applications, cloud services, and SaaS platforms directly accessible to agentic AI. Traditional APIs are being supplanted by Model Context Protocols (MCPs), a highly expressive protocol that lets AI agents spin out child agents and interact deeply with internal systems. While MCPs were designed for functionality, they currently lack essential security, safety, privacy, accountability, and compliance controls, creating new trust‑boundary vulnerabilities that are attractive to attackers. Rosenquist argues that by 2026 the industry must harden MCPs, govern the growing “shadow AI” ecosystem, and implement robust controls; otherwise, we risk cascading cyber‑attacks with catastrophic consequences.

Chapters

The widespread and rapid adoption of AI increases the overall cyberattack surface. The immediate risk we face with AI adoption is not so much the new tools themselves, but rather with enabling current enterprise solutions to be AI accessible. As Agentic AI momentum grows in value and capabilities, global enterprise applications, cloud services, and SaaS offerings will want to be accessible by these new agents. The traditional way to connect software is to use APIs, application programming interfaces, but those are rudimentary compared to what agentic systems require. So they are being replaced with a new protocol, MCPs, model context protocols. These are very expressive, informative, flexible, and can spin out child agents to go deeper into other internal systems. The protocol was designed for functionality to empower connectivity with the Agentic AI systems, but have little in the way of security, safety, privacy, accountability, or compliance control. Integration of AI systems and connections to services via APIs and MCPs creates new vulnerabilities and trust boundary issues. They are a cyber attacker's dream and ripe for manipulation, misuse, or compromise. In 2026, the industry must harden MCPs so they are not easily exploitable. We must understand the risks of agentic AI systems and effectively govern the growing shadow AI problem. Otherwise, we will witness cyberattacks that cause cascading failures leading to catastrophic impacts.