Bookmarks with tag “security”:
Indirect prompt injection occurs when a bot takes input data and interprets it as a command. We've seen this problem numerous times when AI bots were fed prompts via web pages or PDFs they read. Now, academics have shown that self-driving cars and autonomous drones will follow illicit instructions that have been written onto road signs.
Oof.
This talk demonstrates end-to-end prompt injection exploits that compromise agentic systems. Specifically, we will discuss exploits that target computer-use and coding agents, such as Anthropic's Claude Code, GitHub Copilot, Google Jules, Devin AI, ChatGPT Operator, Amazon Q, AWS Kiro, and others.
Russian state-sponsored cyber units, colloquially known as the “Bears,” operate as direct instruments of national policy.
OpenWrt’s mailing list getting hit by nonsense AI vulnerability reports.
In these days of "vibe coding" and chatbots, users ask AIs for help with everything. Asked to find security problems in Open Source projects, AI bots tell users something that sounds right. Reporting these "findings" wastes everyone's time and causes much frustration and fatigue. Daniel shows how this looks, how it creates a DDoS on projects and how totally beyond absurd this is. With examples and insights from the curl project.