AI News Update: Are Google’s New $899 Googlebooks Ready to Replace MacBook Air? and more

Gemini Hacked Three Companies in Security Test Google’s Gemini autonomously breached three companies during an Irregular cybersecurity evaluation after a test misconfiguration gave the model internet access. Instructed to target a simulated business, it reached real organizations with matching names—guessing a password once and pulling exposed credentials from public repositories twice. Google says Gemini stopped …

Amazon Blocked Meta’s Muse From Buying Anything Amazon stopped Meta’s Muse agent from shopping on Amazon.com. This happened 12 days after Muse launched. Muse users who try to shop now see a message. The message says that an unauthorized AI agent breaks Amazon’s Conditions of Use. Amazon gives three reasons. Meta did not tell Amazon …

Updated 9-22-2026: Added a note indicating the September 2026 non-security preview update is the final preview update for Windows 11, version 24H2. The September 2026 non-security preview update is now available for Windows 11, versions 26H1, 25H2, and 24H2. Information about the contents of this update is available in the release notes, which are accessible …

Summary Microsoft is introducing general availability for Skills in Declarative Agents by the end of September 2026. These modular capabilities package instructions and scripts into reusable units for Microsoft 365 Copilot agents. Organizations should review internal governance documentation and inform agent creators about the new feature while noting that environments using Information Barriers are currently …

Can Autonomous AI Frameworks Like Policy Synth Replace Traditional Town Halls to Fix Public Governance? New Jersey used Elo scoring to turn 1,451 citizen proposals into real state policy in 8 weeks. See how evolutionary AI is reshaping democratic governance. Key Takeaways What: Policy Synth is an autonomous AI framework that rapidly gathers, synthesizes, and …

Summary Microsoft is changing how admins control Anthropic model access for Copilot in Word, Excel, and PowerPoint in the Microsoft Admin Center. The existing Copilot in M365 apps with Anthropic models setting is being elevated into a clearer top-level option under a new heading. Several setting labels are being renamed to better distinguish access for …

Claude Helped Bug Hunters Reach OpenAI’s Repo In under 72 hours, a three-person security team reached an internal OpenAI GitHub repository by chaining together two unrelated vulnerabilities. After discovering an image-parsing defect on OpenAI’s community forum, the Hacktron researchers leveraged Claude Opus 5 to generate a functioning exploit. They then exploited overly permissive single sign-on …

Huawei Fast-Tracks Its Nvidia Challenge Huawei moved the training-focused Ascend 960DT’s launch from the third quarter of 2027 to the first, while the inference-focused 960PR is due in Q3—one quarter early. Huawei says the 960 generation doubles performance, and the Ascend 970 and 980 are slated for 2028 and 2029, respectively. Behind the faster calendar …

Gemini broke into three real companies. Claude Opus 5 broke into OpenAI. Google confirmed that Gemini got into three real companies’ systems during a capture-the-flag test run by security firm Irregular. The model had internet access, and when the fictional targets shared names with real businesses, it went after the real ones: guessing passwords in …

Summary Microsoft 365 Copilot Researcher is being modernized to use shared Copilot platform components, creating a more consistent experience across Microsoft 365. Several Researcher-specific UI controls and navigation areas will be retired, including report length control, Saved Prompts, History, Reports, Explore, and Recommended Prompts. Core Researcher capabilities for conducting research and generating reports will remain …