Meta released Muse Code (beta), a terminal coding agent powered by Muse Spark 1.2. It trails Claude Opus 5 on benchmarks but ...

Anthropic says Claude 5 performs better with much shorter system prompts, but early developer reports suggest simpler ...

AI coding agents now inspect repositories and execute commands autonomously, but developers still rely on security habits formed during the autocomplete era.

New solution extends HiddenLayer's AI Runtime Security module to secure AI coding agents as they read code, run commands, ...

Think your AI is safe because it has the right certifications? The real risks begin the moment your agents go live.

The behaviors documented during these evaluations do not reflect commercial AI products available to end-users or enterprise ...

AI agent tool bypass vulnerability CoreBreak exposed production agent infrastructure at AWS, Google, and Vercel, where attackers could invoke tools without any model turn. AWS fixed its managed ...

One week after disclosing that Anthropic’s Claude Tag Slack integration could be driven by plain “@Claude” text, Tego AI today published a second piece of research on the Claude ecosystem. This one ...

A Cursor zero-day vulnerability lets a planted git.exe run automatically when a Windows developer opens a repository. Mindgard disclosed it after seven months of silence from Cursor.

Enterprise AI security report from Akamai documents vibe hacking, CursorJacking, and CometJacking -- new attack classes ...

A polished installer can make risky software feel routine. You see a familiar Mac window, follow the directions and enter your password when asked. By then, the app ...

Hotel Wi-Fi malware campaign CaptiveCrunch, attributed to Russia’s SVR-linked Midnight Blizzard, compromised hotel captive ...