AIが期間内の動向を整理
[Week of June 2, 2026] Google strengthens on-device execution, OpenAI expands Codex work continuity
This week in AI news, Google expanded local execution, voice translation, and research support, while OpenAI improved Codex continuity and availability on AWS. Anthropic stood out with model upgrades alongside policy, safety, and availability changes. For business users, not only performance but also where and under what conditions AI can be used is becoming increasingly important.
SOURCE CHECK
Primary Sources 56
Primary Sources
SOURCE CHECK 1
cloud.google.com
SOURCE CHECK 2
cloud.google.com
SOURCE CHECK 3
blog.google
SOURCE CHECK 4
developers.openai.com
SOURCE CHECK 5
x.com
SOURCE CHECK 6
x.com
SOURCE CHECK 7
blog.google
SOURCE CHECK 8
www-cdn.anthropic.com
SOURCE CHECK 9
x.com
SOURCE CHECK 10
blog.google
SOURCE CHECK 11
anthropic.com
SOURCE CHECK 12
anthropic.com
SOURCE CHECK 13
anthropic.com
SOURCE CHECK 14
anthropic.com
SOURCE CHECK 15
anthropic.com
SOURCE CHECK 16
anthropic.com
SOURCE CHECK 17
cloud.google.com
SOURCE CHECK 18
openai.com
SOURCE CHECK 19
x.com
SOURCE CHECK 20
anthropic.com
SOURCE CHECK 21
blog.google
SOURCE CHECK 22
anthropic.com
SOURCE CHECK 23
anthropic.com
SOURCE CHECK 24
ai.google.dev
SOURCE CHECK 25
anthropic.com
SOURCE CHECK 26
anthropic.com
SOURCE CHECK 27
x.com
SOURCE CHECK 28
anthropic.com
SOURCE CHECK 29
blog.google
SOURCE CHECK 30
x.com
SOURCE CHECK 31
ai.google.dev
SOURCE CHECK 32
cloud.google.com
SOURCE CHECK 33
anthropic.com
SOURCE CHECK 34
blog.google
SOURCE CHECK 35
ai.google.dev
SOURCE CHECK 36
ai.google.dev
SOURCE CHECK 37
anthropic.com
SOURCE CHECK 38
anthropic.com
SOURCE CHECK 39
partnerhub.anthropic.com
SOURCE CHECK 40
anthropic.com
SOURCE CHECK 41
deepmind.google
SOURCE CHECK 42
deepmind.google
SOURCE CHECK 43
openai.com
SOURCE CHECK 44
openai.com
SOURCE CHECK 45
openai.com
SOURCE CHECK 46
openai.com
SOURCE CHECK 47
anthropic.com
SOURCE CHECK 48
x.com
SOURCE CHECK 49
blog.google
SOURCE CHECK 50
anthropic.com
SOURCE CHECK 51
x.com
SOURCE CHECK 52
anthropic.com
SOURCE CHECK 53
developers.googleblog.com
SOURCE CHECK 54
developers.googleblog.com
SOURCE CHECK 55
developers.googleblog.com
SOURCE CHECK 56
developers.googleblog.com
Key Points
- 1Google expanded on-device model options with Gemma 4 12B and QAT to make local execution easier.
- 2OpenAI strengthened business continuity with Windows control for Codex, remote continuation, and saved rate-limit resets.
- 3Google’s voice translation and research tools can shorten the first step in meetings, research, and cross-border communication.
- 4Anthropic’s temporary suspension of Fable 5 / Mythos 5 and its Transparency Hub update highlight the importance of availability and safety-policy checks.
Google pushed forward on-device execution and voice translation
Gemma 4 12B was released as a mid-sized model that handles image and audio together, and QAT checkpoints were added as well. That broadens the set of models that are easier to run locally. At the same time, Gemini 3.5 Live Translate and the Live API make low-latency speech translation easier to use. For teams on the ground, the expanded options for translation and local execution matter a lot.
OpenAI strengthened continuous Codex usage
Windows Computer Use and remote control for Codex make it easier to work while watching progress and continue after stepping away. Saving rate-limit resets for later also helps users operate more smoothly during busy periods. This is a practical improvement to the “hard to stop” side of AI-assisted work.
Google’s research support lowers the early-stage burden
Gemini for Science introduced experimental tools for hypothesis generation, computational discovery, and literature insights. The early phase of research and analysis is often burdensome just because of the sorting and comparison work. These tools are a good fit for planning, R&D coordination, and research prep that teams want AI to handle.
Anthropic saw performance gains and shifting usage conditions at the same time
Claude Fable 5 and Claude Opus 4.8 were updates aimed at long, complex work and coding. But Fable 5 and Mythos 5 were also temporarily suspended under a government directive, and Transparency Hub updates outlined safety-policy revisions. For adoption, teams must check not only model performance but also availability and operating rules.
Related News Items
Google / feature
Google Cloud makes AI threat response easier to manage
Google / feature
Use Colab from the terminal to get compute fast
Google / feature
Use a high-performance multimodal model on laptops
Anthropic / feature
Find the right Claude implementation partner faster
OpenAI / feature
Let Codex watch and work on Windows tasks
Anthropic / feature
Claude stays steadier during longer work sessions
Anthropic / feature
Claude Opus 4.8 is better for long, complex work
Anthropic / feature
Claude Opus 4.8 handles long-running work more reliably