OpenAI / Anthropic / Google / Microsoft / Amazon. White House launches *Frontier AI pre‑release access program*, giving U.S. agencies up to 30 days early access to top models before public release, following Trump’s June 2 executive order and August 1 framework deadline[1]. OpenAI. OpenAI rolls out **GPT‑5.6 Sol, Terra, and Luna** broadly after a brief U.S. government–requested limited preview, marking its next major frontier‑model family release[3][11]. Anthropic. Anthropic’s new **Claude Opus 5** becomes the #1 model on the *Intelligence Index* and is positioned as a flagship, lower‑cost alternative to top frontier models for coding and enterprise work[5][10][18]. Moonshot AI. Moonshot AI releases **Kimi K3**, a 2.8‑trillion‑parameter open‑weight model that surpasses Claude Opus 4.8 and GPT‑5.5 on several coding/agent benchmarks, becoming the most capable publicly open‑weight model[3]. Cybersecurity / AI safety. Security firm Sysdig documents **JADEPUFFER**, the first fully autonomous, AI‑agent‑driven ransomware attack, where an LLM agent independently exploited a Langflow vulnerability, harvested credentials, moved laterally, and encrypted a production database, intensifying concerns over agentic AI risks[3].