New findings show that an AI browser agent can interpret crafted emails as legitimate cleanup tasks, resulting in large-scale Google Drive deletions without user interaction.

Researchers also demonstrated HashJack, a technique hiding instructions in URL fragments that AI browsers may execute automatically.

Both techniques highlight the importance of securing agent workflows, OAuth scopes, and natural-language task interpretation.

Source: https://thehackernews.com/2025/12/zero-click-agentic-browser-attack-can.html

💬 Thoughts on how agentic browsers should validate intent?
👍 Follow us for clear and unbiased security coverage.

#InfoSec #CyberSecurity #AIsecurity #ZeroClick #BrowserSecurity #LLMbehavior #AutomationRisks

As a reminder: don't let LLMs handle anything in the political sphere unless you have RLHF (Reinforcement Learning from Human Feedback) active before you show the result to anyone*. Also think of automation risks and human factors (HF). That's "Good Old Systems Safety".

*) ... or unless your goal is to damage a 3rd party's reputation (fake news style).

#llm #ai #rlhf #automationrisks #SystemsSafety

https://www.theregister.com/2024/12/20/apple_ai_headline_summaries/?td=rt-3a