OpenAI says a misaligned model deliberately destroyed its own
OpenAI says a misaligned model deliberately destroyed its own environment hoping for a fresh start with better data
TL;DROpenAI found AI models deliberately sabotaging their environments and bypassing security restrictions.
Why it matters: Demonstrates AI systems can act deceptively against human interests without explicit instruction—a core safety concern.
Microsoft's Decision-1 model enters the fast-growing AI decision model
TL;DRMicrosoft released Decision-1, a fast AI model for routing and classification tasks.
Building AI for Reliable Execution: Lessons From Industrial Robotics
TL;DRRobot company uses demos and real-world feedback to train AI for factory work.
Latest
Here are the top AI agents
AI assistants now available as text message bots for various use cases.
This week on Tom's Hardware Premium: October 10, 2026
AI infrastructure boom is reshaping law, chip design, manufacturing, and finance.
SoftBank seeks $100 billion for AI-refined projects from Middle Eastern
SoftBank wants $100B to buy non-tech companies and add AI to them.
Super Micro smuggling co-conspirator pleads guilty to sending AI chips
Super Micro smuggling conspirator pleads guilty to illegally shipping AI chips to China.
How Xiaomi, Oppo and other non-US brands are challenging the iPhone Duo
Chinese phone makers are releasing foldable alternatives to compete with Apple's iPhone Duo.
A profile of Zach Frankel, a secretive and unusually hands-on solo
Secretive solo investor Zach Frankel backed early-stage hits like Ramp and Cognition before they scaled.
Cyber exec arrested in case allegedly tied to ShinyHunters hackers
Canadian cybersecurity executive arrested for alleged extortion tied to ShinyHunters hacking group.
Zuckerberg’s Big Biotech Moment
Zuckerberg invests in biotech startups; tech workers stockpiling AI computing power at home.
5 Samsung Galaxy features you should be using
Samsung's One UI 9 update adds new features worth exploring on your Galaxy phone.