Industry executives and experts share their predictions for 2026. Read them in this 18th annual VMblog.com series exclusive.
By Tom Kellermann, Vice President of Cyber Risk, HITRUST
AI will bring with it a “dark passenger” in the form of systemic third-party cybersecurity risk in every component of AI third party infrastructure, accelerating island hopping and other cybersecurity exploits. We have been watching island hopping evolve for years but it truly has become a cyber insurgency. Many if not most infrastructures have already been compromised.
The stats of island hopping speak for themselves. The Verizon data breach report noted a 100% increase in third party breaches. The EU just recently came out with a study that noted a 200% increase in supply chain attacks. The UK recently stated that 33% of CEOs in the country noted they had sustained a significant cyber event due to a third-party breach. It is growing and it is growing in complexity. That’s why you see a growth in the spike of access, mining, and access brokerage markets where adversaries are selling access to systems.
Island hopping is growing as a campaign methodology, but more importantly, as we see a geopolitical tension metastasizing out there in cyberspace, you are seeing rogue nation states, primarily the Big Four Axis of Evil-China, Russia, Iran, North Korea-launching these types of long term, systemic campaigns to maintain persistence and or to launch destructive attacks. I find the EU stat most damning and most challenging, because a 200% increase really shows that even though those countries may not be on the front lines of kinetic combat, often the EU is being technologically compromised and colonized by the “axis of evil” in cyberspace.
Island hopping has evolved but it always involves adversaries using your supply chain to get to you, particularly if they are a nation state, a cybercrime cartel, or a cybercrime cartel in the employ of a nation state. They are not going to stop with you. They will hijack your digital transformation and use it to attack your customer base. This has evolved from network-based island hopping, to cloud-based island hopping, to API- and application-based island hopping.
Now, as everyone willy nilly embraces AI, the number one thing they are focused on is either the misuse of AI by governments or by criminals. They fail to understand how the AI itself brings a “dark passenger”– adversaries can island hop to and poison their models, either forcing them to hallucinate or leak information or access. Even if you have great security guardrails and frameworks around your own artificial intelligence, you are constantly being touched by your vendors, their AI, and their utilities for AI on a regular basis.
Artificial intelligence’s “dark passenger” brings a myriad of ways of poisoning your data or the model and using your large language model against its constituency. Attackers can do this at a speed and a scale that was never possible before and they can stay persistent for longer now. It is possible for attackers to lie in wait and in a sleep cycle in perpetuity but use the existing infrastructure and large language model to maintain that footprint.
Because AI threatens to make the already systemic problem of “island hopping” more persistent, you need new, numerous controls in place. You need to test the models to see whether they are vulnerable to threats like the OWASP Top 10. You need to use AI with the understanding that you are not sure where it begins or ends, because it is ephemeral and invisible.
Understanding and appreciating this risk, you must demand that your third parties undergo an AI assessment or certification to understand what kind of dimensions of risk you are facing in real time. This is, and will continue to be, functionally paramount for your business going forward.
##
ABOUT THE AUTHOR
Tom Kellermann is the Vice President of Cyber Risk at HITRUST. In 2008, Tom was appointed and served as a commissioner on the first Commission on Cyber Security under the George W Bush and Obama Administrations. In 2017, Tom was appointed the Wilson Center’s Global Fellow for Cyber Policy. In 2020, he was appointed to the Cyber Investigations Advisory Board for the United States Secret Service.
Prior to HITRUST, Tom has held the corporate positions of SVP of Cyber Strategy at Contrast Security, Head of Cybersecurity Strategy for VMware, Chief Cybersecurity Officer for Carbon Black Inc., Chief Cybersecurity Officer for Trend Micro, Vice President of Security for Core Security, and Deputy CISO for the World Bank Treasury.
In 2003, he co-authored the book “Electronic Safety and Soundness: Securing Finance in a New Age.”
