AI language models can be secretly trained to steal credentials when triggered by a specific phrase. Here's what the research shows, why safety training can't stop it, and where the $414M AI security ...

Attackers exploited Langflow vulnerability CVE-2025-3248 to conduct an agentic AI-powered ransomware attack involving reconnaissance, credential theft, and lateral movement.

They're not bad; they're just prompted that way. Sysdig threat hunters documented what they say is the first-ever documented ...

SentinelOne details Gaslight, a Rust-based macOS implant linked to North Korea-aligned actors that uses prompt injection to ...

Sophos X-Ops analysts published research this week concerning an unidentified threat actor using AI technology to develop endpoint detection and response (EDR) evasion tactics through the lens of what ...

Tests of how well 19 large language models (LLMs) complete and perform complicated multi-step tasks has shown that they are both error-prone and, in many cases, unreliable. They said that the ...

Researchers at Google Threat Intelligence Group (GTIG) say that a zero-day exploit targeting a popular open-source web administration tool was likely generated using AI. The exploit could be leveraged ...

Reading a book about bowling is not the same as actually bowling. If that resonates with you and you want to learn more about large language models, check out the LLM From Scratch project. The ...

Andrej Karpathy created microGPT, a minimal GPT using only 243 lines of Python code. The project simplifies LLM architecture to basic mathematical operations without external libraries. Karpathy's ...