Daily AI RoundupsMonday, 22 June 2026

22 June was about putting AI into production. Anthropic updated Claude's identity verification for access, trust and abuse controls, and Samsung Electronics deployed ChatGPT Enterprise and Codex worldwide in a major OpenAI enterprise rollout. Sakana said its Fugu model matches Fable and Mythos benchmarks and routes around vendor restrictions, then shipped Fugu Ultra to Vercel's AI Gateway with one-to-three-agent routing. On infrastructure, NVIDIA detailed Rubin AI servers running at 45C with 100% liquid cooling for efficiency. Open-source tooling kept pace: llama.cpp b9745 added Step3.5/3.7 flash MTP3 and speculative draft paths, and PyTorch patched a bmm Triton launch bug hitting vLLM tests across CUDA devices.

• Enterprise AI
• Model releases
• AI infrastructure
• Dev tooling
• Trust & safety

Stories in this roundup

1. 01
   Claude Support updates identity verification on June 21 for access, trust and abuse controlsANTHROPIC
2. 02
   Sakana says Fugu matches Fable, Mythos benchmarks and routes around vendor restrictionsSAKANA AI
3. 03
   NVIDIA's Rubin AI infrastructure uses 45C, 100% liquid cooling to improve AI server efficiencyNVIDIA
4. 04
   Samsung Electronics deploys ChatGPT Enterprise and Codex worldwide in major OpenAI enterprise rolloutOPENAI
5. 05
   llama.cpp b9745 adds Step3.5/3.7 flash MTP3 support and speculative multi-head draft pathsGITHUB
6. 06
   Sakana AI’s Fugu Ultra hits Vercel AI Gateway, routing work to 1-3 agentsVERCEL
7. 07
   PyTorch patches bmm Triton launch bug hitting vLLM tests when CUDA tensors sit on cuda:1 but device is cuda:0GITHUB