At the RSAC 2026 cybersecurity conference, major vendors like CrowdStrike and Microsoft introduced new frameworks to manage the growing identity risks of AI agents.

A CMS misconfiguration exposed Anthropic's unreleased Claude Mythos model, which the company confirmed marks a 'step change' in reasoning, coding, and cybersecurity.

AI experts warn deepfake fraud has gone industrial with cheap, accessible tools enabling personalized scams. UK consumers lost £9.4bn in 9 months.

Britain collaborates with Microsoft to develop deepfake detection system. 8 million deepfakes shared in 2025, up from 500K in 2023.

OpenClaw, the open-source AI agent that can autonomously manage emails, browse the web, and control your computer, has exploded in popularity but security experts warn of exposed credentials, prompt injection attacks, and hundreds of malicious packages.