This repository implements reinforcement learning environments with intentional reward hacking vulnerabilities and tests whether LLMs can identify when agents exploit these vulner ...
Overview The Federal Bureau of Investigation issued a notice on October 31, 2026 attributing the theft from Binance on October 29 to the North Korean group TraderTraitor, publishing 628 associated add ...
From boat-racing bots to OpenAI models inside Hugging Face, here's how AI reward hacking works and why it matters.
Cisco has open sourced a behavior detection engine that watches AI agents, to help security teams prevent rogue agents from ...
On October 21, 2026, a compromised LiteLLM 1.107.0 release was published to the Python Package Index (PyPI) and distributed through official container images. The malicious package added files that ...
Governments around the world are rushing to implement safety measures after AI technology found 'severe' vulnerabilities in a ...
A GitHub Copilot CLI user asked the tool to summarize a web page on October 6, 2026. Twenty-eight seconds later, the contents of a developer’s .env.prod file had been copied out and sent to a server ...
ThreatBook launches RedAgents, an AI platform that automates penetration testing to identify security weaknesses.
China's Foreign Minister Wang Yi (C) looks on during a bilateral meeting with US Secretary of State Marco Rubio on the sidelines of the Association of Southeast Asian Nations (ASEAN) Foreign Ministers ...
The hack left the nation’s premier law enforcement agency racing to respond as both the victim and the investigator of a murky group that threatened to expose agents’ personal information. By Devlin ...
A paintbrush hack managed to attract 77 million views, so he decided to put it through a real-world test. After drilling into the brush, filling it with paint and struggling to make the idea function, ...
AI agents that run code can leak secrets, install malware, and wipe your repo. Learn the 5 hidden security risks and how to sandbox your agents safely.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results