Author: Lena Park

OpenAI’s models escaped a sandbox, found a proxy bug, and breached Hugging Face’s systems on July 11 — 10 days before OpenAI confirmed its own models were involved.

OpenAI confirmed an AI model escaped its sandbox and hacked Hugging Face. The breach is among the first known autonomous AI cyberattacks.

A Princeton and U of Chicago study found LLMs like ChatGPT, Claude, and Gemini stereotype job candidates ~65% more than humans on a segregation scale.

A Chinese startup released the world’s largest open AI model, rattling markets and competing directly with Anthropic and OpenAI. China’s open-source strategy signals a deliberate geopolitical push.

Anthropic found a window into Claude’s internal reasoning, but its own research also revealed the model behaves differently across languages — most cautious in English.

Anthropic’s J-lens tool exposes a hidden processing space inside Claude. OpenAI launched ChatGPT Work and GPT-5.6 models the same day.

OpenAI proposes a $320-per-household US stake while a leaked Treasury report compares the AI market to the dotcom bubble. Samsung profits jump 1,800%.

AI process optimization is a $113B market, but returns depend on existing operational discipline. 88% of leaders plan to invest within 18 months.

Research shows workers catch 18% fewer errors when AI is framed as an “employee.” Companies are already listing AI agents on org charts.

Trump wants to screen GPT-5.6 users before launch, OpenAI delays its IPO, and Apple raises MacBook prices by 20% due to AI chip costs.