Researchers Use Poetry to Jailbreak AI Models Summary Researchers from Sapienza University of Rome, the Sant’Anna School of Advanced Studies and LLM safety consultancy Dexai published a white paper showing…
Major AI conference flooded with peer reviews written fully by AI Article Date: 27 November 2025 Article URL: https://www.nature.com/articles/d41586-025-03506-6 Article Image: https://media.nature.com/lw767/magazine-assets/d41586-025-03506-6/d41586-025-03506-6_51767010.jpg Summary Pangram Labs analysed text from submissions and…
Large language models are biased — local initiatives are fighting for change Article Date: 2025-06-26 Article URL: https://www.nature.com/articles/d41586-025-03891-y Article Image: https://media.nature.com/lw767/magazine-assets/d41586-025-03891-y/d41586-025-03891-y_51698972.jpg Summary Global large language models (LLMs) such as GPT,…
How Malware Authors Are Incorporating LLMs to Evade Detection Summary Google’s Threat Intelligence Group (GTIG) and reporting from Dark Reading outline how threat actors are experimenting with large language models…
Building compositional tasks with shared neural subspaces Article Date: 26 November 2025 Article URL: https://www.nature.com/articles/s41586-025-09805-2 Article Image: Figure 1 Summary This study trained two rhesus macaques to switch between three…
Slipknot-gauged mechanical transmission and robotic operation Summary This paper introduces a simple, purely mechanical method to encode and transmit force information using slipknots tied in suture or filament — a…
Vicarious body maps bridge vision and touch in the human brain Summary Article Date = 26 November 2025 Article URL = https://www.nature.com/articles/s41586-025-09796-0 Article Title = Vicarious body maps bridge vision…
Prompt Injections Loom Large Over ChatGPT’s Atlas Browser Summary OpenAI’s new Atlas browser brings ChatGPT agentic capabilities into a Chromium-based web browser, making autonomous agents available to a much wider…
HashJack attack shows AI browsers can be fooled with a simple ‘#’ Summary Cato Networks has disclosed “HashJack”, a novel indirect prompt-injection technique that hides malicious instructions in the URL…
ShadowRay 2.0 Turns AI Clusters into Crypto Botnets Summary Threat actors are exploiting a known remote code execution flaw in the open-source Ray distributed computing framework to take over AI…
