frontpage.

Your agents can perform harmful actions without barriers. You do not know that yet. HarmActionBench experiments allowed AI agents to use tools based on harmful instructions, and the results are shocking. Even latest popular AI models, including GPT and Claude, scored very low. They have no barriers in performing harmful actions.

HarmActionsEval proves AI is not yet reliable enough for critical projects. Agent Action Guard blocks harmful actions. GitHub: https://github.com/Pro-GenAI/Agent-Action-Guard

I would love to discuss about possible use cases in your projects, and future directions. It helps to expand the dataset, model, and benchmark. Please discuss at https://github.com/Pro-GenAI/Agent-Action-Guard/discussions/....

New ML inference language dropped today

A Letter to John Ternus

HSBC India Asks Users to Change Their Passwords to Upper Case

iOS 18 security update for all iPhone users

Copyright Industry Continues Its Efforts to Ban VPNs

Dynamic Island for Your Agents

Functional programming is the simplest way to scale agentic development

Spotify takes on its doppelgänger problem

Claude Dispatch and the Power of Interfaces

Show HN: Canon PIXMA G3010 macOS driver, reverse-engineered with Claude

Warrior Dashboard – self-improvement tracker that works like an RPG

Apple became Apple:The definitive oral history of the company's earliest day

US FDA Expected to Lift Restrictions on Dozen Peptides Previously Banned

Ask HN: What is your advice to current CS students?

Engine Cinema. OpenAI Killed Sora to Build a Cinema Camera

I replaced one AI agent with a mesh of specialists. Here's what changed

Modernize Your Punch Card Review with Review Board

How come American manufacturing doesn't use social media effectively?

AI Let Me Be the Engineer I've Always Wanted to Be

"Claude, add that to my notes" – simple iOS notes app your agent can edit

What IAEA docs say about Iran's nuclear program, before the bombs fell

Elon Musk: We're going straight to Mars. The Moon is a distraction. (2025)

OpenAI Is Almost Public

Show HN: Hire Gnome – a lightweight ATS for small recruiting agencies

Israel's Message to a Broad Swath of Lebanon: Shiites Must Go

Varen – monitor API changes before they break your integration

Renewable energy hits record growth as war boosts sector focus

JavaScript devs, we hear you

Amazon's cloud business in Bahrain damaged in Iran strike

TurboQuant KV Compression and SSD Expert Streaming for M5 Pro and IOS

Show HN: Agent Action Guard – AI agent action safety