AIが期間内の動向を整理
OpenAI, Anthropic, and Google at the end of June: AI for work is becoming more practical
The June 30 AI news was dominated by announcements with clear workplace relevance: OpenAI’s new option for security work, Claude becoming easier to use in Slack, Claude’s general availability on Azure, and Gemini’s screen-operation support. For people who cannot track news every day, it is worth understanding what changed and why it matters.
SOURCE CHECK
Primary Sources 8
Primary Sources
Key Points
- 1OpenAI previewed GPT-5.6 Sol with strong security-task performance
- 2Anthropic released Claude Tag to make Claude easier to call from Slack
- 3Claude became generally available in Microsoft Foundry, improving Azure usage
- 4Google DeepMind highlighted computer use in Gemini 3.5 Flash to expand screen-task automation
OpenAI expands options for security work
GPT-5.6 Sol was introduced as a model with strengths in vulnerability research and command-line tasks. Security work often involves long hours of verification, so a model focused on this area can help make the initial phases of investigation and triage more efficient.
Claude fits more naturally into Slack workflows
Claude Tag is notable because it allows requests to be handed over directly from Slack conversations. Since only selected channels and tools can be connected, it is easier to start with a limited scope and use it for team research and follow-up task organization.
General availability on Azure directly affects adoption decisions
With Claude now generally available in Microsoft Foundry, Azure-based organizations can handle authentication and billing in one place. Because it fits existing operational and management workflows, this matters not just for experiments but for production adoption decisions.
Gemini’s computer use is tied to automating repetitive tasks
Computer use in Gemini 3.5 Flash broadens the ways models can handle on-screen interactions. Availability is still limited, but organizations that want to reduce repetitive web and app work should treat this as an important signal for future workflow design.