Hello!

Welcome to Secure Prompt’s weekly newsletter, issue #6.

From Shadow Escape zero-click attacks compromising MCP-based AI assistants to Google’s unresolved ASCII smuggling flaw in Gemini, this week underscored how rapidly AI security is evolving - and breaking. Chatbots are being weaponized as backdoors, OpenAI’s new Guardrails can already be bypassed, and AI-powered malvertising campaigns are spreading phishing links at scale. Meanwhile, nearly three-quarters of UK workers admit to using “Shadow AI,” leaving CISOs scrambling to regain control.