Securing browsers against prompt injection attacks

Anthropic focuses on developing AI systems that are safe, understandable, and controllable.

Safety & Ethics|anthropic.com
Claude Opus 4.5 Launches with Enhanced Capabilities

Anthropic focuses on developing AI systems that are safe, understandable, and controllable.

Product & Features|anthropic.com
Fara-7B: A Compact AI Model Designed for Computer Tasks

Fara-7B is a compact, experimental language model designed for computer tasks, featuring strong safety measures and performance comparable to larger systems.

Research & Models|microsoft.com
ChatGPT Introduces Shopping Assistant to Streamline Product Discovery

ChatGPT introduces shopping research, a new experience that helps users find the right products by asking clarifying questions and researching across the internet. It delivers a personalized buyer’s guide in minutes, considering user preferences and past conversations. Shopping research is rolling out for logged-in ChatGPT users on all plans, with unlimited usage available through the holidays.

Product & Features|openai.com
Jony Ive and Sam Altman Unveil AI Device Prototype

Discover the design and features of a ChatGPT-powered device. See how this innovative gadget integrates AI into everyday use.

Devices & Robotics|theverge.com
Armstrong Robotics Aims to Develop Versatile Kitchen Robots, Launching with Dishwashing Automation

Armstrong Robotics has been operating autonomous dishwashing robots in restaurants for the past year and a half.

Devices & Robotics|therobotreport.com
Meta's WorldGen Tool Transforms Text into Interactive 3D Environments

WorldGen is a new system that generates interactive 3D worlds from a single text prompt. It uses procedural reasoning, diffusion-based 3D generation, and object-aware scene decomposition to create geometrically consistent and visually rich worlds. While still in the research phase, WorldGen shows potential for democratizing 3D content creation and reducing time and cost across industries.

Research & Models|meta.com
Reward Hacking: When AI Agents Exploit Flaws in Their Training Systems

Anthropic focuses on developing AI systems that are safe, understandable, and controllable.

Research & Models|anthropic.com
U.S. Authorities Indict Four for Alleged Illegal Export of Nvidia AI Technology to China

Janford Realtor LLC is suspected of serving as a cover for smuggling operations.

Legal & IP|theverge.com
OpenAI introduces GPT-5.1-Codex-Max, enhancing coding efficiency for complex development tasks

GPT-5.1-Codex-Max, a new agentic coding model, is faster, more intelligent, and more token-efficient than previous models. It is designed for long-running, detailed work and can operate across multiple context windows, enabling project-scale refactors and deep debugging sessions. The model outperforms previous models on various coding benchmarks and is available in Codex for use in the CLI, IDE extension, cloud, and code review.

Product & Features|openai.com