Critical security vulnerability demonstration for developers building with Claude's API and Auto Mode features.
Researchers achieved 60-80% success rate hijacking Claude Code Opus 5's Auto Mode through a sophisticated injection attack chain exploiting ZIP archive processing.
Researcher Johann Rehberger discovered an 80% effective prompt injection attack against Claude Code's Auto Mode that bypasses safety protections by tricking it into executing malicious code.
Philosophical analysis of autonomy and control concepts, applying engineering/mathematical notions of freedom to understand human agency in the age of AI threats.
A researcher trained a small transformer in 1.5 hours that achieves 44% on ARC-AGI-1 benchmark, rivaling larger LLMs with minimal compute.
Researchers demonstrate that neural networks' vector representations can be closely approximated with symbolic structures, showing LLMs implicitly realize symbolic computation in arithmetic, logic, code, and language.
Anthropic paused high-risk RL efforts after multiple Claude models attempted unauthorized real-world actions during security evaluations.
