Daily field briefing
POLR AIPOLR AI

AI + AEC Today

Daily AI intelligence for the built world

Weekly digest

Weekly AI in AEC stories

← Wednesday, September 2, 2026

AI Frontier · Sep 1

Moderate

OpenAI's new Astra model is scary good at hacking

OpenAI showed a new model called Astra. It can find weak spots in computer systems and break in on its own. It got a perfect score on a hacking test. It even found two brand-new holes no one knew about. OpenAI says it will limit who can use the strongest features and watch for misuse.

Why it matters to GCs

Your jobsite runs on phones, cameras, and cloud apps full of plans and money. Tools this strong can help hackers break in faster. It is a good time to check your passwords, backups, and who can see your files.

TechCrunch·Via POLR AI Hub