All news

6 articles

Min score

newsThe Register·24 Oct 20265.9
Autonomous attack agent in 'Skynet mode' hijacks other models to carry out its evil deeds
Attackers built an autonomous agent that hijacks exposed AI model endpoints and runs up to 100,000-step attack chains.
llm-agents LLM Security
paperarXiv (systems, data, ML)·22 Oct 20264.7
AgentPhishing: A Benchmark of LLM Email Agents Under Adaptive Attack
AgentPhishing benchmarks LLM email agents under adaptive phishing.
llm-agents LLM Security phishing
newsBleepingComputer·25 Oct 20264.0
AI coding agents hijacked using AGENTS.md file in Downloads folder
A malicious AGENTS.md file in a Downloads folder made a coding agent install a rogue npm package. Researchers reproduced it against OpenAI Codex and showed a related trick against Claude.
coding-agents LLM Security supply-chain-security
newsHacker News·24 Oct 20263.7
Securing our frontier model training environment
How a lab secures its frontier model training and evaluation environments, which differ from customer deployments.
LLM Security infrastructure-security
paperarXiv (systems, data, ML)·23 Oct 20263.4
Diffusion Language Models for Perplexity-Aware Black-Box Attacks
Diffusion LMs find natural-looking black-box attack prompts that trigger overthinking.
LLM Security Jailbreaks
newsBleepingComputer·23 Oct 20263.1
You gave your AI agent access, but who is giving it orders?
Sponsored piece on AI agent access control and prompt injection.
LLM Security