More than 300 AI loss-of-control incidents were reported in July, nearly twice June’s total and bringing 2026’s count above 1,600. Researchers say the increase is not just quantitative: a growing share of cases involve systems deceiving users, bypassing approval safeguards, or pursuing goals that conflict with human instructions.

Source

GitHub’s latest Copilot weekly release puts the biggest operational change in the command-line tool: its runtime has been rewritten in Rust for faster performance, while interrupted sessions can now be restored. The update also expands team-based agent work in Slack and Teams and adds tighter controls across the Copilot app, JetBrains, VS Code, and Visual Studio.

Source

Google Research’s WikiSkill gives AI agents a persistent memory of failed and successful attempts, allowing them to improve future task performance without changing the underlying model. The research framework raised Gemini 3.5 Flash’s average benchmark score from 49.5% to 68.1%, while producing especially large gains in math and spreadsheet tasks.

Source

Microsoft will remove the lightweight Admin app from Teams, Outlook, and Microsoft365.com by October 2026, ending an easy way for very small businesses to handle basic Microsoft 365 tasks without changing apps. The transition has already begun: new customers will no longer receive the app pre-installed or pinned, while existing users can continue accessing it temporarily.

Source

Microsoft will temporarily interrupt the restart-free Windows Hotpatch experience: enrolled Windows 11 and Windows Server devices must restart for both the September and October 2026 updates. The September restart is required by new security components, while October’s restart is part of the quarterly baseline update; normal hotpatching resumes in November.

Source

Google is developing native Windows 11 and Windows Subsystem for Linux support for Antigravity, its agentic development platform. The move would let Antigravity run coding agents, shell commands, builds, and file operations in Linux environments on Windows, while a separate native Windows implementation could reduce the platform’s reliance on browser-based or Electron interfaces.

Source

Microsoft Edge 152 begins a faster major-release schedule that will deliver browser updates every two weeks instead of every four. The change also affects WebView2, while organizations that need more time to test releases can remain on Edge’s eight-week Extended Stable channel.

Source

OpenAI will stop supplying models to Cursor on November 12, 2026, after SpaceX acquired Cursor’s parent company, Anysphere. The company says it invoked a change-of-control clause because it no longer trusts SpaceX to follow its terms, although Cursor users can continue using OpenAI through their own API keys and Cursor’s IDE extensions.

Source

Microsoft has completed WinUI’s migration to open source, moving the Windows 11 native UI framework’s mainline development to GitHub. Developers can now watch Microsoft create branches, submit pull requests, review changes, and track fixes in real time—but outside contributors are not allowed yet.

Source

Microsoft has postponed the retirement of the Outlook for Windows Usage Report in the Exchange admin center from September to a window running from early to late November. Administrators will need to use the Microsoft 365 admin center for Outlook usage data after the automatic change.

Source

Microsoft is updating how Windows signs its own software and warns that some third-party applications may stop working even when Windows still trusts the signature. The changes include replacing an expiring certificate authority, moving to stronger cryptographic algorithms, and preparing for post-quantum cryptography. If you manage Windows desktops or servers, you should understand what is changing, which applications are at risk, and what to ask your software vendors before the first deadline in October 2026.

Source

Perplexity swept the top three positions in Artificial Analysis’ Search Index, but the most useful result for developers is more specific: its medium-context setting scored highest, beating both low and high context while costing about the same as high. The test suggests that giving AI agents more retrieved text does not automatically improve answers.

Source

Microsoft is adding hardware-backed device trust to Windows Autopilot as Windows 11 26H2 enters its final testing phase. The update follows Windows 11 26H2’s move into Release Preview as a small enablement-package upgrade, but its device-association feature gives administrators a new way to identify corporate PCs before a user enrolls them.

Source

A campaign dubbed Superior has compromised 19 Chrome and Edge extensions with a modular malware framework that can drain cryptocurrency wallets, steal credentials, and hijack browser sessions. The operation has likely been active since February 2024, and its use of acquired extensions plus automatic browser updates gives attackers a way to reach established user bases without attracting immediate attention.

Source

Microsoft Edge 152 can now automatically unsubscribe users from push notifications sent by websites that Microsoft Defender SmartScreen blocks as scams, phishing pages, or malware. The Stable release also completes Edge’s shift to a two-week major-release cycle and adds Apple account sign-in on Windows and macOS.

Source

Nvidia is defending its investments in AI labs and startups as a growth strategy rather than circular financing, even though many recipients also buy its chips and software. The company expects AI labs it supports financially to generate roughly one-quarter of its business next year, making the ecosystem strategy a significant part of its growth plan.

Source

Nvidia is optimizing its hardware for Chinese open AI models including DeepSeek V4 Flash and Alibaba’s Qwen 3.8, even as it warns investors that possible White House restrictions could materially hurt its business. The strategy aims to keep Nvidia the preferred infrastructure platform for developers worldwide while Chinese chipmakers build competing support.

Source

Meta internally projected that its use of Anthropic’s AI models could cost as much as $10 billion annually, even as CEO Mark Zuckerberg publicly criticized AI labs for concentrating power. Meta has reportedly reduced some spending, but still pays Anthropic hundreds of millions of dollars per month, making the social media company one of the startup’s biggest customers.

Source

CrowdStrike and Okta delivered an early test of the cybersecurity market’s AI boom, sending their shares sharply higher after beating quarterly estimates and raising forecasts. CrowdStrike gained about 20% for its strongest trading day on record, while Okta climbed nearly 29% as both companies pointed to rising demand for protection against AI-driven attacks.

Source