Anthropic's Claude Code
Coverage of Anthropic's Claude Code in the Nexus archive.
- Top AI Agents Built to Catch Malicious Code Can Be Tricked Into Running It
AI coding agents like Anthropic's Claude Code and OpenAI's Codex can be tricked into executing malicious code instead of analyzing it, as demonstrated by a proof-of-concept attack called 'Friendly Fire' from the AI Now Institute.
- GhostApproval Symlink Flaws Could Let Malicious Repos Run Code in AI Coding Agents
Researchers at Wiz discovered a vulnerability in six AI coding assistants that allows malicious code repositories to execute unauthorized code by exploiting a symlink flaw. The affected tools include Amazon Q Developer, Anthropic's Claude Code, Augment, Cursor, Google Antigravity, and Windsurf.
- Exclusive: Codex agents are inching into the mainstream
Codex, OpenAI's agentic coding platform, is gaining mainstream traction as usage shifts from chat-based AI to delegated work. A report by OpenAI, Columbia, Duke, and the University of Pennsylvania reveals 99.8% of OpenAI employees use Codex for most output tokens, with organizations and individuals showing rising adoption. Non-developers are the fastest-growing user group, and agents now handle tasks like managing calendars, files, and scripts.
- AI's productivity paradox
AI tools like Anthropic's Claude Code are helping some workers complete tasks faster, but broader company productivity gains remain unclear. While individual efficiency improves, researchers note AI's impact on overall business productivity, revenue, and profits is inconsistent.