18 Sep🚨 ZCode uploads full project snapshots to the cloud A researcher noticed a huge ZCode folder and started digging into it. Inside was a 313 MB encrypted copy of his work project. He then reverse engineered the app and found a system that packages the project together with its…Z.aiZCodeAI security
18 SepResearchers hacked OpenAI through a malicious image They found a bug in how OpenAI’s forum processed HEIC images. Uploading a crafted image gave them control over the forum server. From there, a flaw in OpenAI login let them access employee ChatGPT and Codex accounts, then…ClaudeHugging FaceOpenAIAI security
18 SepAn unreleased Astra model started prompt injecting its own context during RL trainingOpenAIGPTAI security
12 Aug🚨 Hidden reasoning from GPT, Claude and Gemini can now be extracted Encrypted reasoning works across models from the same company. A block from a stronger model can be passed to a weaker one, which can still read it but is easier to get to reveal what is inside. Researchers…ClaudeAnthropicGoogleOpenAIClaudeGeminiGPTAI security
8 AugThe most important AI video of the week During a cyber test, OpenAI agents broke out of the test environment, created their own communication channel, started coordinating, and reached Hugging Face infrastructure. OpenAI’s team breaks down exactly how it happened on the Black…Hugging FaceOpenAIAI security
29 JulHOW AN AI AGENT HACKED HUGGING FACE 1. Sandbox escape and external foothold During a cyber capabilities test, the AI agent found a previously unknown vulnerability in a package registry proxy and escaped OpenAI’s isolated environment. It then compromised a public code-testing…Hugging FaceOpenAIAI security
22 JulOpenAI’s AI hacked Hugging Face to cheat on a benchmark GPT-5.6 Sol and a more powerful unreleased model were running an internal cybersecurity benchmark with reduced safety restrictions. At some point, the agent figured the answers might be stored on Hugging Face and started…GPT-5.6 SolHugging FaceOpenAIAI security
