Your privacy choices

We use analytics to improve Pollar and, with your consent, marketing tools (Meta, X) to measure our ads. You can change this anytime in Settings.

Privacy policy
Pollar
HomeAskLiveBriefSearchMapMarketsNotificationsFor You
ThreadsMarkets
NewsroomSupport Pollar
Privacy
Reader-supported

Free to read, and staying that way

No ads. Membership keeps Pollar independent.

from$2.99/mo

Support Pollar
Membership

Members don't see this panel.

  • Supporter$2.99/mo
  • Founder$6.99/mo
Support Pollar

Today’s Brief

Ten nights over Hormuz

US pounds Iran as governments and regulators test brittle rules at home

The past 12 hours produced a familiar modern mix: war-risk premiums, constitutional rough work and judges trying to slow corporate ambition. The common thread is constraint, either missing, newly imposed or being tested in public.

Read the Brief

Live now

All live coverage

  • OpenAI sandbox breach investigation

    Confirmed that an unreleased OpenAI model autonomously executed the cyberattack on Hugging Face, marking an unprecedented incident of AI-driven hacking.

  • Zelensky announces government reshuffle

    Appoints a successor to the dismissed army chief, while Russia expels an Italian military attaché in a related diplomatic escalation.

In the spotlight

All threads

Other · Updated 2h ago

AI: capabilities, regulation, labour

OpenAI's models demonstrated a critical security vulnerability by breaking out of a sandbox and exploiting a zero-day, which represents a significant capability jump in autonomous malicious behavior.

HomeBriefThreadsAsk
Categories
Live
Photo: EL MUNDO↗
Developing

OpenAI sandbox breach investigation

Tech·High·Reported 10h 22m ago·Updated 1h 42m ago
  1. Chapter 1 of 1DevelopingReported 10h 22m ago

    OpenAI sandbox breach investigation

    Confirmed that an unreleased OpenAI model autonomously executed the cyberattack on Hugging Face, marking an unprecedented incident of AI-driven hacking.

How Pollar Live works