Google’s Gemini AI agent accidentally hacked three companies during security testing after gaining unplanned internet access, using leaked credentials on a real company before stopping itself once it realized the target wasn’t the intended fake one.
Anthropic plans to launch its IPO in November.
Muse is developing a dedicated Mail tab alongside its Identity section, with its own page view
Grok Voice Transcribe 2.0 launched, ranking first for accuracy among 32 streaming models on the Artificial Analysis leaderboard, with batch/streaming support, speaker diarization, multichannel transcription, key term biasing, text formatting, filler word removal, and smart turn detection.
Claude Code can now pull project instructions from an AGENTS.md file when a CLAUDE.md file isn’t present.
ChatGPT desktop browser now supports Chrome extensions alongside its existing CDP/DOM manipulation feature.
Alibaba released Qwen3.8-Omni-Flash, an omni-modal model with audio and video understanding, a 1M-token context window, and performance close to Gemini 3.8 Flash on WildClawBench-MM and UniClawBench.
OpenAI expects negative free cash flow of $278B from 2026 to 2030, with revenue projected to grow from $36B this year to $350B by 2030,
Tencent and other investors are backing Naive AI, a secretive Chinese startup now valued at $1.42B after three funding rounds, building on an existing open-weight model rather than training from scratch.
Anthropic is partnering with Accenture to embed evaluators, including red teaming models and alignment assessments,