Digital assistants are becoming trusted, capable intermediaries, which inadvertently makes them a powerful new vector for social engineering. This article examines how attackers could exploit an assistant’s knowledge, influence, and ability to act on a user’s behalf.
TrendAI™ Research examines the unique dynamics and factors that influence the adoption of AI in state-aligned APT activity. The report looks into four countries (China, Russia, North Korea, and Iran) to understand why their state-aligned actors adopt AI in the ways that they do.
We developed Scarecrow, a tool that hides refusal instructions in documents so AI agents ingesting them are prompted to stop processing. Our findings show that this same technique can help defenders disrupt AI-driven access to sensitive documents.
Generative AI now writes clinical notes, drafts patient replies, and decides claims, often faster than healthcare’s oversight can keep pace. TrendAI™ Research maps the resulting “custody gap” and the risk it creates, from the exam room to the back office, where AI-generated artifacts can enter patient records.
Digital assistants are becoming trusted, capable intermediaries, which inadvertently makes them a powerful new vector for social engineering. This article examines how attackers could exploit an assistant’s knowledge, influence, and ability to act on a user’s behalf.
TrendAI™ Research examines the unique dynamics and factors that influence the adoption of AI in state-aligned APT activity. The report looks into four countries (China, Russia, North Korea, and Iran) to understand why their state-aligned actors adopt AI in the ways that they do.
We developed Scarecrow, a tool that hides refusal instructions in documents so AI agents ingesting them are prompted to stop processing. Our findings show that this same technique can help defenders disrupt AI-driven access to sensitive documents.
Generative AI now writes clinical notes, drafts patient replies, and decides claims, often faster than healthcare’s oversight can keep pace. TrendAI™ Research maps the resulting “custody gap” and the risk it creates, from the exam room to the back office, where AI-generated artifacts can enter patient records.
In this report, we present the security and governance issues enterprises should consider when adopting and using AI agents. We lay out the risk structure unique to AI agents and present a security framework and a method for technical and organizational controls to support AI adoption.
Our TrendAI™ Research team developed AIxploit, a framework that systematically maps an agent’s AI-native attack surface through attack testing.
Cómo la IA está acelerando las fallas de seguridad tradicionales y por qué las amenazas conocidas ahora son más difíciles de detectar y contener.
We examine how OpenClaw’s exposed deployments, plaintext credential storage, prompt injection risks, unguarded MCP integrations, and malicious third-party skills amplify the security risks of agentic AI at enterprise scale.
The critical systems we can’t afford to lose are the same ones attackers are racing to target. With a new generation of AI-augmented cyber threats, what used to be weeks of warning can collapse into mere hours. What is the current state of the threat landscape for essential industries, and what can organizations do to fortify their defenses? Uncover this and more in our report.
China-aligned operational relay box (ORB) infrastructure is harvesting open sensor data at scale, and most defenders can't see it. Our report dives into how these ORBs can use open telemetry for data collection and other purposes.
14 de octubre de 2026
10 min