Wisr CEO CEO to Speak at 2025 Information Technology Symposium on Agentic AI and Third-Party Risk Management
Security research shows an AI‑trained malware (using Qwen 2.5 LLM) can bypass Microsoft Defender ~8% of the time, while SRAM-trained variants struggle to reach 1%. Developed in just three months on a $1,600 budget, it demonstrates how reinforcement‑learning enables automated creation of evasive malware at scale. With Black Hat 2025 slated to expose its inner workings, this marks a major escalation in cybersecurity: adversaries are now using AI to automate and optimize attack craft. Responding demands higher fidelity detection models and adversarial ML defenses.
