Coding assistants

CrowdStrike's Falcon AI Detection and Response (AIDR) now extends its AI visibility, detection, and response to Microsoft Copilot Studio and Claude Code, plus a Falcon browser extension capability. In Copilot Studio, Falcon AIDR runs as an external threat detection provider and returns allow or block decisions on tool calls before they execute. For Claude Code, it connects to the tool's hook event system to check prompts and tool activity, with setup done by adding a JSON block to the settings file.

Microsoft is building an AI "super app" that combines Copilot's chat, coding, and agentic capabilities. CEO Satya Nadella confirmed on an earnings call that it will span consumer and commercial experiences when it launches this year.

OpenAI announced GPT-5.6, its latest flagship model series, which will become the new preferred model in Microsoft 365 Copilot across Word, Excel, PowerPoint, Chat and Cowork. Microsoft will serve the models natively and also access OpenAI models directly through the API to bring GPT-5.6 to Microsoft 365 customers.

Microsoft has fixed a known issue that removed the Copilot Chat or Copilot buttons from Classic Outlook for Windows users holding the Copilot Chat (Basic) license. Microsoft also says it is investigating Outlook crashes on systems running Kaspersky Antivirus, linked to the Kaspersky Mail Checker (mcou.dll).

A critical, three-stage attack against Copilot, dubbed 'SearchLeak', enables one-click data theft. The attack is now patched and belongs to a new group of AI prompt-injection issues that use hidden URLs and other variables.

OpenAI announced it will acquire Ona, a startup providing secure, pre-configured cloud environments where AI agents can access tools, systems and context. The deal will let OpenAI's coding assistant Codex take on longer-running tasks and help more organizations deploy agents into production. Terms were not disclosed and the deal remains subject to customary closing conditions.

A Checkmarx survey of 2,350 CISOs, AppSec managers and developers across 14 countries found that enterprises with 81% to 100% AI-generated code ship vulnerable code 3.4 times more often than those using AI for 20% or less of their code. About 30% of respondents said they ship compromised code and hope the vulnerability is not found, and 93% reported at least one breach from in-house apps. Only 22% of organizations have formal AI governance.

Apple's WWDC keynote largely matched AI features already available elsewhere, including Siri capabilities found on Android and in the Claude and ChatGPT apps. The article's author then tried the first developer beta of iPadOS 26, and the excerpt ends before describing what that experience showed.

On 2 June 2026, Microsoft announced two text LLMs. MAI-Thinking-1 is a 35B-parameter reasoning model available to select early partners, and MAI-Code-1-Flash is a 5B-parameter model built for GitHub Copilot and VS Code, rolling out to individual Copilot users in Visual Studio Code. Microsoft says both were trained on clean, commercially licensed data, and that MAI-Thinking-1 was trained without distillation from third-party models.

Microsoft is rolling out a redesigned Microsoft 365 Copilot across desktop and mobile, which the company says loads twice as fast. The update adds a feature it calls "progressive disclosure", which shows tools and controls based on the user's prompt rather than all at once.

Google is adding Gemini, the AI assistant, across more of its apps at a fast pace, and the author finds the spread increasingly irritating. The piece compares this to Microsoft's placement of Copilot shortcuts throughout Windows 11, which drew similar user complaints. The source text is truncated and ends before the full story.

Microsoft Edge is adding a feature that lets its Copilot chatbot gather information from all open tabs. Users can ask Copilot about tab contents, compare products and summarize open articles. Microsoft says users can choose which experiences to enable, and it is retiring Copilot Mode, which offered agentic features such as booking reservations.

Simon Willison describes how his view of vibe coding and agentic engineering has begun to converge in his own work, which he calls a disturbing realization. He distinguishes vibe coding, where the person does not look at the code and it is suited to personal tools, from agentic engineering, where a professional engineer uses AI tools while maintaining quality, security and operational standards. He admits he no longer reviews every line of code his coding agents write, even for production systems, and raises the question of whether that is responsible.