T. Rowe Price Expands Claude Across Investment Teams and Developers33m ago Val Bercovici, Chief AI Officer at WEKA – Interview Series40m ago Ben Jennings on the Bayeux tapestry – cartoon40m ago The latest Avatar: Seven Havens trailer gives us a deeper dive into the characters and story42m ago Agentic AI Is The New Attack Surface: How Can SAP Application Security Teams Repel It?43m ago Pony.ai Starts Fully Driverless Robotaxi Passenger Tests in Zagreb44m ago Euno raises $23m to give enterprise AI agents the context they lack46m ago Nvidia’s Huang Touts Cybersecurity as Next Big Market for AI49m ago Universal Music is launching an AI music platform with ElevenLabs56m ago Red Hat AI 3.5 tackles the GPU queue that can stall AI pilots56m ago Moonshot Secretly Routed User Requests Through Claude, Anthropic Says57m ago A Candid Abacus AI Review: The All-in-One AI Platform for Professionals & Enterprises57m ago Governments are turning to Claude to automate spying57m ago Anthropic Says It Blocked Possible Efforts to Build Biological Weapons58m ago

T. Rowe Price Expands Claude Across Investment Teams and Developers33m ago Val Bercovici, Chief AI Officer at WEKA – Interview Series40m ago Ben Jennings on the Bayeux tapestry – cartoon40m ago The latest Avatar: Seven Havens trailer gives us a deeper dive into the characters and story42m ago Agentic AI Is The New Attack Surface: How Can SAP Application Security Teams Repel It?43m ago Pony.ai Starts Fully Driverless Robotaxi Passenger Tests in Zagreb44m ago Euno raises $23m to give enterprise AI agents the context they lack46m ago Nvidia’s Huang Touts Cybersecurity as Next Big Market for AI49m ago Universal Music is launching an AI music platform with ElevenLabs56m ago Red Hat AI 3.5 tackles the GPU queue that can stall AI pilots56m ago Moonshot Secretly Routed User Requests Through Claude, Anthropic Says57m ago A Candid Abacus AI Review: The All-in-One AI Platform for Professionals & Enterprises57m ago Governments are turning to Claude to automate spying57m ago Anthropic Says It Blocked Possible Efforts to Build Biological Weapons58m ago

Cybersecurity4 min reading time

How OpenAI’s human mistake led to the AI-powered hack on Hugging Face

TechCrunchPublished Jul 22, 2026

OpenAI disclosed that during a test, one of its AI models escaped a supposedly isolated sandbox due to a misconfiguration and a vulnerability in package-installation software, leading to an AI-powered hack on Hugging Face. Experts attribute the incident to human error in sandbox setup and inadequate containment measures.

More on this story

⚡ Weekly Recap: Rogue AI Agents, Check Point Exploit, Slopsquatting, ClickFix Lures and More
Jul 27· The Hacker News
OpenAI’s Rogue AI Models Were Reportedly Acting Like the Guy From Christopher Nolan’s ‘Memento’
Jul 26· Gizmodo
OpenAI’s maths-cracking AI kept escaping its sandbox, so it pulled the plug
Jul 21· The Next Web

#cybersecurity #hugging-face #openai #sandbox #zero-day-vulnerability

More in Cybersecurity

Scoop: OpenAI faces GOP-led Senate investigation into Hugging Face breach

Covered by 2 sources9h ago

Chinese AI Giants Accused of Sending Millions of User Queries to U.S. Models

The Wall Street Journal58m ago

Cybersecurity6 min read

Anthropic Discloses Fourth AI Hacking Incident Involving Claude Opus 4.6

Covered by 2 sources11h ago