Microsoft has released version 1.0 of its open-source Agent Framework, positioning it as the production-ready evolution of the project introduced in October 2025 by combining Semantic Kernel ...

What if AI doesn't actually engage in thought at all? An advanced AI model known as Centaur astonished researchers by matching human performance across 160 cognitive evaluations. However, a recent ...

AI Mode in the Google app on Android only currently supports image uploads. File uploads only work right now in AI Mode’s web interface. That may soon change, with the Google app getting support for ...

Students from Atlanta Georgia's Spelman College recently spoke to the press about their 'PlantGPT' project. The system picks up telemetry on humidity, light intensity, soil moisture and outside ...

Data and artificial intelligence literacy skills are no longer considered specialized, a new report concluded, and instead have “crossed a critical threshold” as something widely expected in the ...

Misconfigurations and access control issues trigger many enterprises security problems, according to IBM’s X-Force security research and threat intelligence group. AI tools allow attackers to identify ...

Posts from this topic will be added to your daily email digest and your homepage feed. is an investigations editor and feature writer covering technology and the people who make, use, and are affected ...

Many journalists are using artificial intelligence. But best practices have not been widely codified, and are limited in what they address, judging by a briefing from the Center for News Technology & ...

Researchers at Stanford and Caltech have found some critical reasoning failures in advanced AI models. LLMs are great at recognizing patterns, but they have trouble with basic logic, social reasoning, ...

Investigation of 30 top AI agents finds just four have published formal safety and evaluation documents relating to the actual bots. The new wave of AI web browser agents, many designed to mimic human ...

Here’s what you’ll learn when you read this story: Large language models (LLMs) like ChatGPT show reasoning errors across many domains. Identifying vulnerabilities is good for public safety, industry, ...