AI News Summary 2026-06-12
.. lang: en
AI News — June 12, 2026
GAFAM and Major AI Companies
DiffusionGemma Is No Longer Just a Paper or a Demo: It’s Entering the Real-World Stack
- Google published the DiffusionGemma guide, vLLM announced native support, and NVIDIA presented it as optimized for RTX, RTX PRO, and DGX Spark.
OpenAI Brings Codex Closer to Enterprise Infrastructure
- OpenAI announced that eligible Oracle Cloud customers can use OCI commitments for OpenAI models and Codex. It also announced its intention to acquire Ona to enhance Codex with persistent, customer-controlled cloud environments.
Anthropic Introduces Access-Controlled Capabilities
- Anthropic launched Claude Fable 5 and Claude Mythos 5. Fable 5 is widely available; Mythos 5 is limited to trusted users in sensitive domains.
Influencers and Tech Blogs
The technical conversation around Claude is shifting toward governance
- On June 11, Simon Willison highlighted a governance angle regarding Claude and policies for researchers, as documented in his daily log. The confirmed release remains Claude Fable 5/Mythos 5.
Hugging Face Provides Insights into Applied Engineering for Voice Agents
- The Hugging Face blog covered recent technical activity, including an analysis of code-switching in ASR for voice agents.
Generative Imaging
Canva Expands Generative Imaging into Editing and Production
- Canva announced that Magic Layers works within Gemini and ChatGPT to convert AI-generated images into editable Canva designs.
Chatbots and Agents
Codex Emerges as a Persistent Business Agent
- OpenAI announced access to models and Codex via Oracle Cloud and a plan to acquire Ona, cloud infrastructure for long-running Codex agents.
JumpCloud Puts Identity and Auditing at the Heart of Agents
- JumpCloud Agentic IAM launched on Google Cloud, optimized for Gemini Enterprise, with a focus on discovery, registration, governance, and auditing of agent identities.
Cresta Targets the Lifecycle of Agents in Production
- Cresta Conductor promises to help build, test, and improve AI agents using real customer conversations; more context is available on the Cresta blog.
Anthropic links Frontier agents with risk-based availability
- Claude Fable 5 and Mythos 5 separate general availability from restricted access for sensitive capabilities.
On-Premises AI and Serving
vLLM Begins Serving Text Models via Diffusion
- vLLM added native support for DiffusionGemma. Google documented the model in its developer guide, and NVIDIA published its approach to local acceleration on RTX/PRO/DGX Spark.
llama.cpp continues to see frequent updates
- The llama.cpp releases show automated activity on June 11 and 12, with several recent builds and fixes.