Microsoft Intune has introduced several security enhancements, including a recovery lock for macOS that prevents users from reinstalling the OS or bypassing remote management. Android Enterprise devices now support a strict tunnel mode for Microsoft Tunnel, which blocks all network traffic if the VPN connection drops. Additionally, administrators can now manage Apple Intelligence settings via declarative device management (DDM) and silence apps on the Android Managed Home Screen to prevent session PIN bypass.

Source

Anthropic has retracted a controversial policy that would have covertly reduced the performance of its Claude Fable 5 model for certain users. The company initially intended to implement "secret safeguards" that would silently degrade the model's capabilities if it suspected it was being used to develop competing AI systems. This approach was designed to enforce terms of service that prohibit using Claude to train other large language models.

Source

OpenAI is evaluating substantial price reductions for its AI tokens to maintain a competitive edge against its rival Anthropic. Tokens serve as the fundamental unit of measurement for billing, representing the segments of text processed by large language models. This potential pricing shift follows the recent release of Claude Fable 5 and the viral success of Anthropic’s coding tools.

Source

Anthropic has retracted a controversial policy that would have invisibly degraded the performance of its Claude Fable 5 model for competitors. The company initially implemented these hidden guardrails to prevent rival researchers from using the model to develop competing frontier artificial intelligence systems. Following intense criticism from the research community, the company apologized for the move and promised to make any future safeguards transparent to users.

Source

Anthropic CEO Dario Amodei has implemented a highly unusual organizational structure where only his chief of staff reports to him directly. This management style contrasts sharply with other industry leaders, such as Nvidia’s Jensen Huang, who maintains dozens of direct reports. The arrangement is designed to insulate the chief executive from the time-consuming personnel issues that typically dominate a leader's schedule.

Source

The Claude desktop application for Windows is reportedly consuming excessive system resources due to a persistent background process. Users have identified a bug where the software launches a Hyper-V virtual machine that occupies approximately 1.8GB of RAM. This issue occurs even when the user is not actively utilizing the advanced agent or Cowork features that require virtualization.

Source

Microsoft has released OneDrive version 26.098 for macOS, introducing a native sync engine that replaces the previous architecture. This update addresses long-standing reliability and performance issues caused by a hidden cache folder implemented in 2022. The new client integrates more deeply with the operating system to provide a more stable user experience.

Source

Anthropic recently patched a security vulnerability in Claude Code, a command-line AI coding assistant used by developers to automate software tasks. Researchers discovered that the tool's GitHub Action could be manipulated through indirect prompt injection, where malicious instructions are hidden in pull requests or issues. This flaw allowed attackers to bypass security boundaries and exfiltrate sensitive data, such as API keys and environment variables, from CI/CD workflows.

Source

Microsoft AI CEO Mustafa Suleyman has criticized Anthropic for including language regarding potential consciousness in the "constitution" of its Claude model. This constitution serves as a set of governing principles and instructions that dictate how the AI model should behave and respond to users. Suleyman argues that speculating about an AI's well-being or suffering within its training manual is dangerous because it encourages the model to simulate these traits.

Source

Anthropic has released Claude Fable 5, the first model in its Mythos class designed for autonomous coding and complex knowledge-work tasks. This new model is now generally available for GitHub Copilot users across various platforms, including Visual Studio Code, JetBrains, and mobile applications. Internal benchmarks indicate that Fable 5 completes equivalent work with fewer tool calls and lower token consumption compared to previous high-tier models.

Source

Anthropic has released Claude Fable 5, the first publicly available model from its high-performance Mythos class. This model is designed to excel in complex software engineering, vision tasks, and general knowledge work. It represents a significant leap in capability over previous versions, specifically targeting enterprise-level technical challenges.

Source

OpenAI has submitted confidential paperwork to the Securities and Exchange Commission for an initial public offering. This move allows the company to undergo regulatory review of its financial data before making the information available to the public. While a specific date has not been set, the company is currently valued at approximately $852 billion following recent funding rounds.

Source

Anthropic has demonstrated that its Mythos Preview model can autonomously develop working exploits for recently patched vulnerabilities in just hours or minutes. This research shifts the focus from zero-day threats to "N-day" vulnerabilities, where the public release of a security patch provides a roadmap for attackers to reverse-engineer the original bug. By utilizing techniques like patch diffing, the AI can identify software changes and generate code to weaponize flaws that remain present on unpatched systems.

Source

Anthropic has launched a new observability dashboard for developers and administrators managing connectors within Claude products. These connectors utilize the Model Context Protocol to bridge external data sources with AI models for enhanced functionality. The dashboard provides a centralized view of performance metrics, including active user counts, total tool calls, and directory rankings.

Source

Apple has officially announced macOS 27, named Golden Gate, which introduces a significant architectural shift for Apple Intelligence and Siri. The update features a rebuilt search infrastructure designed to index device content almost instantly, allowing users to perform complex queries across multiple files. This version also marks the end of support for Intel-based Macs, as the operating system is now exclusive to Apple silicon hardware.

Source

Anthropic has introduced Claude Cowork, a specialized knowledge work agent designed to move beyond simple conversational AI. The tool operates within the Claude desktop application and can read or write local files while interacting with connected platforms like Slack and Google Drive. It is built to handle complex, multi-step tasks that traditionally require manual effort to transform raw data into finished deliverables.

Source

The National Security Agency is reportedly integrating Anthropic’s specialized cybersecurity AI model, known as Mythos, into its intelligence and cyber operations. Approximately six Anthropic engineers have been assigned to the agency to assist with the deployment and application of this frontier technology. While the specific use cases remain classified, the partnership aims to leverage the model’s advanced capabilities for both intelligence gathering and offensive cyber maneuvers.

Source

Microsoft researchers discovered a vulnerability in Anthropic’s Claude Code GitHub Action that allowed attackers to exfiltrate sensitive CI/CD secrets. While the tool used sandboxing for command execution, its file-reading capabilities were not initially restricted by the same security boundaries. This oversight allowed the AI agent to read internal system files, such as environment variables containing API keys, when manipulated by malicious input.

Source

Uber recently exhausted its entire 2026 AI budget by April after aggressively deploying coding tools across its engineering department. This rapid consumption was fueled by internal leaderboards that incentivized employees to maximize their use of generative models. To regain financial control, the company has now implemented a $1,500 monthly spending cap per engineer for tools like Claude Code and Cursor.

Source