Anthropic is investigating unauthorized access to Claude Mythos, a restricted cybersecurity AI model available only to selected companies, cleared organizations, and some government agencies.
Alibaba’s Qwen3.6-Max-Preview is its strongest model yet, with reported benchmark gains, a 256k context window and hosted proprietary access through Alibaba Cloud.
OpenAI launched ChatGPT Images 2.0 on April 21, adding reasoning before generation, multilingual text, real-time information, broader aspect ratios, and up to eight images from one prompt.
Anthropic launched Claude Design, a research preview tool for paid Claude users that turns prompts into prototypes, visual assets, and handoff-ready outputs.
OpenAI loses three senior executives while cutting side projects including Sora and tightening its focus on ChatGPT, the API, and enterprise AI.
Anthropic has released Claude Opus 4.7, its most capable public model, with stronger coding, higher-resolution image support, new safeguards, and unchanged pricing.
OpenAI has launched GPT-Rosalind, a life sciences AI model for biological research, drug discovery, and tool-heavy scientific workflows.
OpenAI has expanded Codex into a broader workplace tool with computer control, memory, plugins, image generation, remote devbox access, and repeatable task automation.
Stanford HAI 2026 AI Index shows AI advancing quickly in reasoning, coding, science, and adoption, while public anxiety and labor disruption intensify.
