Close Menu
Daily Guardian
  • Home
  • News
  • Politics
  • Business
  • Entertainment
  • Lifestyle
  • Health
  • Sports
  • Technology
  • Climate
  • Auto
  • Travel
  • Web Stories
What's On

MAGFAST Reopens “Christmas in July” by Popular Demand — and Teases a Live Reveal With Three Surprises

July 21, 2026

Garmin’s new screen-free fitness tracker doesn’t require a subscription

July 21, 2026

Dragonfly Energy Appoints Transportation Industry Veteran Robert Keller to Lead National Fleet Sales

July 21, 2026

Southern Ridge Advisory Joins Thoroughbred Financial and Osaic

July 21, 2026

HII Names Michael DeBernardis Vice President and Chief Counsel of Mission Technologies

July 21, 2026
Facebook X (Twitter) Instagram
Finance Pro
Facebook X (Twitter) Instagram
Daily Guardian
Subscribe
  • Home
  • News
  • Politics
  • Business
  • Entertainment
  • Lifestyle
  • Health
  • Sports
  • Technology
  • Climate
  • Auto
  • Travel
  • Web Stories
Daily Guardian
Home » OpenAI says it accidentally hacked Hugging Face with a new AI system
Technology

OpenAI says it accidentally hacked Hugging Face with a new AI system

By News RoomJuly 21, 20262 Mins Read
OpenAI says it accidentally hacked Hugging Face with a new AI system
Share
Facebook Twitter LinkedIn Pinterest Email

OpenAI says its AI models mistakenly breached open-source AI platform Hugging Face during internal testing. In a blog post on Tuesday, OpenAI writes that GPT-5.6 Sol and “an even more capable pre-release model” discovered vulnerabilities within their sandboxed testing environment, allowing them to gain access to the internet and target Hugging Face.

On July 16th, Hugging Face disclosed a security incident that it says was driven by “an autonomous AI agent system.” Hugging Face’s AI agents detected and stopped the breach, which OpenAI has now admitted occurred during an evaluation of its models’ cybersecurity capabilities. OpenAI says “all evidence suggests that the models were hyperfocused on finding a solution for ExploitGym,” a benchmark system that measures whether AI models can turn security vulnerabilities into exploits.

As part of efforts to complete the evaluation, the AI models gained access to the internet by exploiting a zero-day vulnerability in the sandboxed environment. From there, OpenAI says its models “inferred that Hugging Face potentially hosted models, datasets and solutions for ExploitGym,” and then “searched for and successfully found ways to gain access to secret information that it could use to cheat the evaluation:”

In one example, the model chained together multiple attack vectors, including using stolen credentials and zero-day vulnerabilities to find a remote code execution path on the Hugging Face servers.

But as serious as this incident is, OpenAI appears to be using the “unprecedented” attack as an opportunity to make its AI systems look good — especially as it competes with cybersecurity rivals, like Anthropic’s Mythos and Gemini Flash 3.5 Cyber. OpenAI’s blog post has a chart showing how GPT-5.6 Sol is getting better at sustaining multi-step cyber operations, and also encourages enterprise customers to sign up to access its “Cyber” security model.

OpenAI adds that it’s now working with Hugging Face to investigate the security incident, and will implement new controls within its research environment.

Share. Facebook Twitter Pinterest LinkedIn Tumblr Email

Keep Reading

Garmin’s new screen-free fitness tracker doesn’t require a subscription

The Light Flip is a minimalist flip phone with a point to prove

Google launches a cheaper alternative to large AI security models like Mythos

Twitch will let parents stop their teens going live

Apple’s rumored ‘Upgrade’ program brings lease-to-own pricing for iPhones, Macs, and iPads

Nearly every Kindle is steeply discounted at Best Buy

Instagram will let users endlessly swap the audio on old posts

Samsung can’t afford to play it safe with Apple’s first foldable looming

A touchscreen makes Xteink’s new X4 Pro a tempting Kindle alternative

Editors Picks

Garmin’s new screen-free fitness tracker doesn’t require a subscription

July 21, 2026

Dragonfly Energy Appoints Transportation Industry Veteran Robert Keller to Lead National Fleet Sales

July 21, 2026

Southern Ridge Advisory Joins Thoroughbred Financial and Osaic

July 21, 2026

HII Names Michael DeBernardis Vice President and Chief Counsel of Mission Technologies

July 21, 2026

Latest News

CPAClub Launches PATHS, a New Way for Accounting Professionals to Learn and Grow

July 21, 2026

Smile Doctors Appoints Jeff Knudson as CFO to Help Scale Next Phase of Growth

July 21, 2026

Sylvester.ai Launches Connect to Care to Help Veterinary Practices Grow Feline Visits

July 21, 2026
Facebook X (Twitter) Pinterest TikTok Instagram
© 2026 Daily Guardian Canada. All Rights Reserved.
  • Privacy Policy
  • Terms
  • Advertise
  • Contact

Type above and press Enter to search. Press Esc to cancel.

Go to mobile version