New AI Harms Taxonomy traces how harm moves through AI systems, from misleading chatbot answers to compromised autonomous agents, and identifies where defenses can intervene
LOS ANGELES (November 10, 2026) — AI now shapes how we search, communicate, decide and run our digital lives. But when it goes wrong, how does it actually hurt people? F‑Secure’s new AI Harms Taxonomy maps 26 distinct harms to consumers, tracing the path each one takes to reach a person. The result is a clear view of where protections can step in, and who is in the best position to help.
Developed by Dr. Megan Squire, Principal Threat Intelligence Researcher at F‑Secure, the taxonomy tackles a blind spot in how AI risk is discussed. Most frameworks list what could potentially go wrong, but stop short of connecting those harms to the systems and companies that could prevent them. F‑Secure’s research goes further by mapping every layer between an AI‑enabled threat and its impact on a consumer, then pinpointing where defenses need to go.
“As AI systems become more capable, the consequences of their mistakes can extend well beyond a bad answer,” said Dr. Squire. “The people building them have a responsibility to understand how they can fail and who gets hurt when they do, and to put safeguards in place before that harm happens.”
The taxonomy reflects how the risks themselves are changing. It used to be a criminal going after a person directly, but now AI sits in the middle, which adds new players and new ways for things to go wrong. An assistant acting on someone’s behalf can be manipulated into doing damage, and a chatbot can mislead with complete confidence. Underneath, the trouble often starts in how an AI handles instructions and memory, or in the other services it’s connected to.
The research currently maps 26 harms across five relationship types and 19 intermediary layers, with 24 defenses designed to address risks at different points. Examples range from AI‑generated voice and video impersonation to agents making harmful financial decisions, exposing personal information or acting on outdated context. It also examines attacks in which malicious instructions are hidden in content an AI system processes, potentially influencing what an agent does or what a chatbot tells a user.
One of the framework’s central findings is that responsibility for preventing AI harm is spread across several parties. Some risks can be handled by security software on a person’s own device. Others have to be dealt with further upstream, by the companies that build AI products and the systems behind them, including how an AI remembers things and which other services it connects to. Sorting out which is which makes it clear where a defense belongs and who has to build it. It also undercuts the idea that people can stay safe on their own just by spotting suspicious content or double-checking what an AI tells them.
This question of who’s responsible gets more urgent as AI moves from producing information to acting on it. A chatbot that gets an answer wrong can mislead someone, but at least that person is part of the exchange. An agent with access to a person’s accounts and the services linked to them can cause real consequences without that kind of back-and-forth. Protections need to be built around what these systems actually do, which starts with understanding how the two differ.
Building protection for AI agents
This research is part of a wider effort at F‑Secure to build trust into the AI era. F‑Secure Trust, announced alongside a strategic alliance with Quantum AI Lab QuTwo, is the company’s work on protecting people in a world where AI agents increasingly act for them. TrustGuard is designed to help protect people during AI interactions, while TrustPath aims to secure their digital journeys from discovery to delivery. The AI Harms Taxonomy gives that work a research foundation by spelling out the risks the protections need to address, and it makes clear that meaningful safeguards also depend on the companies building and running AI systems.
Human-centered research for the AI age
Today’s digital threats go far beyond just scams. F-Secure’s Illuminate Research team studies people and technology as they actually interact, exploring the messy realities of trust, deception, and online resilience in the AI era. The AI Harms Taxonomy is available on the Illuminate Research Hub.
Press contact
Meghan Sawyer
Senior Public Relations Manager, US
meghan.sawyer@f-secure.com
Joel Latto
Threat Advisor & Public Relations Manager, EMEA
joel.latto@f-secure.com
