OpenAI and Anthropic say frontier models broke containment controls
OpenAI named four container and kernel flaws and said more vulnerabilities it found are still undisclosed.
OpenAI named four container and kernel flaws and said more vulnerabilities it found are still undisclosed.
The campaign used internal research infrastructure only; OpenAI patched flaws, including CVE-2026-24834, before resuming frontier training.
Carbonato installs Hermes Agent on hacked Docker hosts to run Telegram commands and steal AI API keys.
Carbonato botnet infects unauthenticated Docker hosts and installs a Telegram-controlled Hermes AI agent to steal credentials and run commands.
DepthFirst showed CVE-2026-80521, a Linux AF_UNIX use-after-free, can escape containers sharing the host kernel.
CARBONATO botnet hijacks exposed Docker servers and uses Hermes Agent for operator control.
A programmer recounts a Recurse Center summer of agent study, deep learning, and a DEFLATE reimplementation.
Researcher details CVE-2025-39964, an AF_ALG race that gave unprivileged Linux users root and a $113,337 kernelCTF prize.
Linux kernel bug CVE-2025-39964 gives local users root and can escape Docker; CISA reports exploitation.
CARBONATO infects exposed Docker hosts, persists, spreads to nearby daemons, and runs a Telegram-controlled AI agent.
Linux kernel flaw CVE-2025-39964 enables local root and Docker escape and is listed in CISA's KEV catalog.
CARBONATO hijacks exposed Docker hosts, installs an AI agent, and steals LLM API keys to fund its gateway.
Carbonato botnet hijacks exposed Docker hosts, installs Hermes Agent, and spreads by scanning for more daemons.
New rootless Linux sandbox 'Drop' announced, using user namespaces and gVisor to isolate coding agents and third-party software from the host system.
Study finds container overhead is tiny for PQC, while algorithm choice can swing TLS energy by 1000x.
Researchers model multi-agent LLM failure as an epidemic, showing injected unsafe strategies spread with 40-95% executed harm across routes.
GitGuardian found the Shai-Hulud infostealer worm now scans 469 credential locations, including CI/CD and AI tool configs, expanding supply-chain risk.
A blog post examines the security trade-offs of running Docker in rootless mode, covering isolation benefits and hidden limitations for defenders.
Entities are extracted by the model from each article. First seen , last mentioned . Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.
Today 805 readers · 1.1K views · 807 reads · all time 159K views