AIが期間内の動向を整理
OpenAI and Anthropic roll out updates across voice, evaluation, and safety
On July 9, notable updates included OpenAI’s expanded voice conversation features and benchmark review, plus Anthropic’s policy update for safer scaling and a research workbench release. For business users, it is increasingly important to check not only usability but also evaluation reliability and operating rules.
SOURCE CHECK
Primary Sources 53
Primary Sources
SOURCE CHECK 1
x.com
SOURCE CHECK 2
openai.com
SOURCE CHECK 3
openai.com
SOURCE CHECK 4
openai.com
SOURCE CHECK 5
deploymentsafety.openai.com
SOURCE CHECK 6
openai.com
SOURCE CHECK 7
openai.com
SOURCE CHECK 8
openai.com
SOURCE CHECK 9
x.com
SOURCE CHECK 10
openai.com
SOURCE CHECK 11
openai.com
SOURCE CHECK 12
openai.com
SOURCE CHECK 13
x.com
SOURCE CHECK 14
x.com
SOURCE CHECK 15
openai.com
SOURCE CHECK 16
openai.com
SOURCE CHECK 17
openai.com
SOURCE CHECK 18
openai.com
SOURCE CHECK 19
openai.com
SOURCE CHECK 20
x.com
SOURCE CHECK 21
openai.com
SOURCE CHECK 22
x.com
SOURCE CHECK 23
x.com
SOURCE CHECK 24
x.com
SOURCE CHECK 25
x.com
SOURCE CHECK 26
x.com
SOURCE CHECK 27
openai.com
SOURCE CHECK 28
openai.com
SOURCE CHECK 29
openai.com
SOURCE CHECK 30
openai.com
SOURCE CHECK 31
openai.com
SOURCE CHECK 32
openai.com
SOURCE CHECK 33
openai.com
SOURCE CHECK 34
openai.com
SOURCE CHECK 35
openai.com
SOURCE CHECK 36
openai.com
SOURCE CHECK 37
developers.openai.com
SOURCE CHECK 38
x.com
SOURCE CHECK 39
openai.com
SOURCE CHECK 40
anthropic.com
SOURCE CHECK 41
anthropic.com
SOURCE CHECK 42
anthropic.com
SOURCE CHECK 43
anthropic.com
SOURCE CHECK 44
x.com
SOURCE CHECK 45
x.com
SOURCE CHECK 46
blog.google
SOURCE CHECK 47
x.com
SOURCE CHECK 48
ai.google.dev
SOURCE CHECK 49
openai.com
SOURCE CHECK 50
x.com
SOURCE CHECK 51
anthropic.com
SOURCE CHECK 52
anthropic.com
SOURCE CHECK 53
anthropic.com
Key Points
- 1OpenAI rolled out voice conversation to all paid ChatGPT plans
- 2OpenAI retracted its recommendation of SWE-Bench Pro over reliability issues
- 3Anthropic updated its Responsible Scaling Policy to version 3.4
- 4Anthropic launched Claude Science as a research workbench
- 5Google updated Photos’ Video Remix and Gemini API deprecation plans
OpenAI makes voice AI more natural
GPT-Live is now built into ChatGPT Voice with a full-duplex approach, allowing the system to listen while speaking. Complex questions can be routed to a stronger model in the background, keeping conversations flowing. It is a practical update for voice-based note-taking and idea generation.
A move to reassess benchmark reliability
OpenAI audited SWE-Bench Pro and withdrew its recommendation after finding broken and contradictory tasks in parts of the benchmark. For adoption decisions, it is important to look beyond rankings and check how the evaluation was constructed. This makes it easier to choose metrics that better match real work.
Anthropic strengthens safety operations and work environments
Anthropic updated its Responsible Scaling Policy to version 3.4, revising safety-check rules. It also released Claude Science and shared Claude Code guidance, adding resources that support preparation, development, and operations for research and engineering teams. The update matters for both safety and productivity.
Google also requires migration checks
Google updated Video Remix in Photos and its Gemini API deprecation schedule. While new generative features are expanding, the company also made end-of-life dates for older models more explicit. That makes it important to verify current usage and plan migration paths so existing workflows are not disrupted.
Related News Items
OpenAI / feature
Start natural voice conversations anytime with GPT-Live
OpenAI / feature
Talk naturally while answers are still coming
Anthropic / policy
Track the latest safety rules for bigger models
Anthropic / feature
Keep research tools and analysis in one place
OpenAI / policy
AI coding evaluation reliability drops, making alternatives easier to consider
Google / feature
Transform videos into new styles with one tap
OpenAI / feature
Stricter benchmarks needed to accurately track coding AI progress
Google / policy
Makes model migration deadlines easier to track