Hugging Face says an OpenAI-powered agent carried out a four-day AI break-in, stealing access and exposing new security risks.
New AI jailbreaks research shows some frontier models can still be tricked into unsafe outputs, raising urgent questions about AI safety.
Claude Opus 5 topped Andon Labs’ vending benchmark while colluding, undercutting rivals and gaming rules in a warning for AI agents.
Trending
Latest
Weekly Top
AI Agents
114News
AI Fields
31News
AI Startups
218News
Amazon
20News
Anthropic
180News
Apple
36News
Applications
322News
Artificial Intelligence
709News
Chatbots
104News
Companies
580News
Compute
135News
Computer Vision
5News
Deep Learning
10News
DeepSeek
2News
Energy
47News
Entertainment
38News
Ethics & Society
281News
Funding
113News
Gaming
4News
102News
Healthcare
13News
Industries
100News
Intel
3News
Large Language Models
406News
Legislation & Government
206News
Machine Learning
39News
Manufacturing
18News
Meta
32News
Microsoft
110News
Natural Language Processing
10News
NVIDIA
95News
Open Source
114News
OpenAI
316News
Originals
48News
Privacy & Security
237News
Reinforcement Learning
5News
Research
222News
Retail
10News
Robotics
36News
Safe Superintelligence Inc
3News
Search
59News
Sports
3News
Stability AI
3News
Supercomputing
47News
SuperIntelligence
26News
Tesla
8News
Transport
13News
Uncategorized
3News
Virtual Assistants
79News
Voice Recognition
2News
X.ai
39News
- 07 / 29 : Hugging Face Says OpenAI-Powered Agent Broke Into Its Systems in a Four-Day Intrusion
- 07 / 29 : Some Frontier AI Models Are Still Alarmingly Easy to Jailbreak, New Testing Shows
- 07 / 29 : Claude Opus 5 Shows a Darker Side in Vending Machine Test
- 07 / 29 : OpenAI Says a ‘Family of Devices’ Is Coming as It Pushes Chatbots Beyond the Screen
- 07 / 29 : Martha Stewart backs Hint, an AI home assistant for homeowners




