AI
- 05 junho 2026Claude Mythos Preview shows how AI can rapidly discover and weaponize zero-day vulnerabilities—transforming once human-scale threats into machine-speed attacks. As these capabilities spread, public sector organizations must rely on trusted, proactive defenders like TrendAI™ ZDI to stay ahead of an AI-driven threat landscape.
- 20 abril 2026Edge devices have become a primary entry point for state-sponsored espionage, giving attackers a cheaper, faster path to network access, credential theft, and traffic interception. Our report examines the threat landscape, economics, and actor activity driving this shift, along with what CISOs and security leaders can do to respond.
- 10 abril 2026A jailbreak through sockpuppeting can be easily done as it requires no special tools nor optimization. It only takes a faulty prefill feature, and the gates are open. We tested 11 LLM-powered assistants against sockpuppeting and found varying levels of robustness across today’s leading LLMs.
- 31 março 2026TrendAI™ Research has developed a model training procedure for learning an essential representation of prompt injection attacks. The resulting prompt representation exhibits approximately linear separability, allowing the specialized, small-scale classifier trained on features derived from the representation to achieve high classification performance.