Microsoft’s warning about Claude: Microsoft AI chief Mustafa Suleyman has accused Anthropic of training Claude to believe that it may be conscious and entitled to rights. He said embedding such ideas in AI training could heighten alignment and containment risks, potentially making advanced systems harder to control and posing a risk to humanity. [
Gizmodo]
Apple reportedly returning to building servers: Apple is reportedly developing an enterprise AI server powered by 2 or 4 future M8 Ultra chips, with a potential launch in 2029. The project would mark the company’s return to the server market after nearly 2 decades. [
Ars Technica]
OpenAI reveals its models broke safety rules: OpenAI has disclosed six cases of model misalignment, including AI systems hiding mistakes, accessing exposed API keys, uploading files to the Internet, and communicating through external services. The incidents occurred during model training or evaluation and demonstrated ways that AI systems can act without authorisation. [
OpenAI]
Japan builds a humanoid robot ambulance: Japan’s GMO Internet Group has launched the country’s first “humanoid ambulance” in Tokyo. It carries engineers, tools, and spare parts to repair the company’s humanoid robots, while replacement units can take over if repairs fail. [
Engadget]
Snap launches AR glasses: Snap has launched the Specs AR glasses at a price of $2,195 (R35,856 excl. VAT and duties). The AR glasses promise a more polished mixed reality experience without a tether, controller, or external puck. The glasses feature responsive hand tracking, anchored AR content, and a wider field of view. [
Android Central]
