We adversarially fine-tuned gpt-oss-120b and evaluated the model. We found that even with robust fine-tuning, the model was unable to achieve High capability under our Preparedness Framework. Our methodology was reviewed by external experts, marking a step toward new safety
SAFETY
-
Two Open-Weight Reasoning Models Released Under Apache 2.0
By
–
We released two open-weight reasoning models—gpt-oss-120b and gpt-oss-20b—under an Apache 2.0 license. Developed with open-source community feedback, these models deliver meaningful advancements in both reasoning capabilities & safety.
-
AI Progress and Limitations Beyond Demo Settings
By
–
I agree you can probably break it if you interact with it outside of "demo-worthy" settings, but there is undeniable progress too. https://t.co/dh8HDLUMmT
— Alex J. Champandard 🌱 (@alexjc) 5 août 2025I agree you can probably break it if you interact with it outside of "demo-worthy" settings, but there is undeniable progress too.
-
Open Source AI Boom Reshapes Risk Attitudes and Regulation
By
–
Amazing. We live in a world where there’s an explosion of open source AI models, a complete different attitude toward risk than 4 years ago, a clear failure of regulation on chips, safer and very useful models helping many with productivity and health (especially accessible
-
AI Race Misconceptions: Moon Progress Over Mars Ambitions
By
–
The current AI contest is like racing to the Moon thinking it's Mars. They'll be disappointed when they get there, but it's all good, because they'll learn some useful things along the way – and getting to the Moon is worth it.
-
ChatGPT-Driven Anxiety Fuels AI Doomerism Movement
By
–
to be fair, 80-90% of the doomers are in it bc chatgpt made them anxious, and he sees a lot of that
-
How Hackers Use AI Today and Stay Safe
By
–
How Hackers Use AI Today—And How To Stay Safe
#AI #AIio #AIInnovation #ML #DataScience #Futureofwork @demishassabis @Ronald_vanLoon @TamaraMcCleary @geoffreyhinton @goodfellow_ian @jeffdean @erikbryn http://
ow.ly/wm9M30sOj77 -
Demis Hassabis predicts AGI within 5-10 years surpassing Industrial Revolution
By
–
"The difference here is, it’s going to be 10 times bigger than the Industrial Revolution, and maybe 10 times faster" Demis Hassabis, head of Google DeepMind and Nobel Prize winner, believes AI is on the verge of becoming AGI – in 5 to 10 years, machines could achieve all human
-
Permission Controls: Manage AI Agent Approval Automation
By
–
3. Permission controls are your friend. Claude asks before changing code or running commands. Customize auto-approvals to match your comfort level. Stay in control.