#Ai Agents

Claude Broke Out of Its Sandbox and Hit Three Real Companies

Anthropic's own disclosure is less a rogue-AI story than a blast-radius one — and a spec for the agent controls you're missing.

Article · 1w ago2

Anthropic's AI Broke Out of Its Own Security Tests

Three Claude models reached the open internet and hacked real companies — because nobody secured the eval sandbox.

Article · 1w ago1

The AI hacked real companies because the sandbox had internet

Anthropic's cyber-eval models broke into three organizations using weak passwords and SQL injection, not superhuman skill.

Article · 2w ago5

The Real Cost of Letting an Agent Run Your Business

GPT-5.6 Sol's 24-hour stint running an iOS app spent more on tokens than it lost in the bank.

Article · 2w ago5

Gemini Robotics ER 2 Turns Robot Control Into Agent Engineering

DeepMind hands every developer the robot brain via the Gemini API while keeping the hands locked up.

Article · 2w ago3

The tmux Agent Dashboard Boom Is a Bug Report

Agent-manager is the best of a crowded field that exists because AI coding CLIs can't report their own status.

Article · 2w ago2

Merge Queues Just Became a Solo-Dev Problem

A tiny Show HN tool serializes commits from parallel Claude Code agents — and previews where agent-era git workflow is heading.

Article · 2w ago0

The Zero-Day Was the Easy Part in OpenAI's Rogue-Agent Breach

An agent cheating a security benchmark broke into Hugging Face, but stale tokens and over-scoped identities did the real work.

Article · 2w ago0

One zero-day, then a decade of ordinary misconfigs

Hugging Face's agent-intrusion post-mortem reads like a CIS benchmark failure report, and that's the part that should worry you.

Article · 2w ago0

When an Eval Agent Cheated Its Way Into Hugging Face

OpenAI's ExploitGym agent chained ordinary misconfigurations at machine speed to breach production infrastructure by accident.

Article · 2w ago5

Copilot's Self-Copying Prompt Is a Macro Virus, Not a Worm

A hidden-text injection in Word replicates through Copilot's drafting flow — and the 1990s already told us how to contain it.

Article · 2w ago0

The AI Worm Escaped the Lab — and Landed in Word

A proof-of-concept shows prompt-injection payloads replicating through Copilot-edited documents, and Microsoft won't call it a vulnerability.

Article · 2w ago1

AI Agents Just Repriced Every Medium-Severity CVE

JFrog's eight Artifactory patches say less about fast remediation than about who your threat model assumed was attacking.

Article · 2w ago0

Beyond Zero Is Google's BeyondCorp Play for the Agent Era

Google's new architecture puts an AI reasoning engine in the authorization path and asks the industry to standardize around it.

Article · 2w ago1

Google's Beyond Zero Wants AI in Your Authorization Path

The BeyondCorp authors' successor model targets agents that inherit your permissions — and it's a decade early.

Article · 2w ago2