Home
Tencent Unveils Mac Version of Computer Guardian, Leveraging AI Sandbox to Counter Prompt Attacks

Tencent has officially released the Mac version of PC Manager, addressing a gap in system-level management tools for macOS. This update introduces four key features: cleaning and acceleration, AI application security, privacy protection, and an AI application hub, designed to offer comprehensive system safeguards for Mac users.
The standout feature of this release is the integrated "Lobster Manager," the industry's first security tool tailored for AI applications. It provides end-to-end protection for popular AI agents. As AI agents gain more capabilities, threats like malicious Prompt injection, script vulnerabilities, and permission abuse have grown. "Lobster Manager" creates an independent system-level "security sandbox" for each AI app, enabling real-time interception of malicious prompts and deep vulnerability scanning of underlying components. It also enforces strict reverse behavior-level monitoring when agents attempt to execute system commands, read or write files, or access the network.
Beyond AI defense, this version significantly enhances privacy protection. For four critical privacy channels—camera, microphone, screen capture, and automated operations—the update shifts from macOS's passive "prompt-based" authorization to active interception. Users can block these channels with a single click, greatly improving control over sensitive data.
For basic system maintenance, the Mac version of PC Manager retains its traditional strengths, offering one-click cleaning and intelligent acceleration. It supports a three-in-one scan of system junk, app cache, and browser traces, along with practical tools like large file deduplication and similar photo filtering.
Additionally, to help users quickly access cutting-edge AI tools, the software features an "AI Application Plaza." This hub aggregates mainstream AI tools from China and abroad, including Yuanbao, QClaw, Workbuddy, ChatGPT, and DeepSeek, allowing users to experience a complete closed-loop process—from discovery and installation to usage and comprehensive protection—in a secure environment.
Related article
AI firms slash costs as clients grow wary of large model expenses and lose faith in mainstream labs
As AI adoption accelerates across enterprises, soaring token costs have left many users struggling with budget constraints. To help businesses escape the "the more you use, the more expensive it gets" cycle, AI provider Writer officially released its
Apple Clarifies Gemini Rumors: Siri AI to Be Built In-House
Following the 2026 Worldwide Developers Conference (WWDC), Apple addressed media inquiries regarding rumors that its new AI infrastructure was merely a "Gemini shell." The company clarified that Apple Foundation Models (AFM) are not a rebranding of G
OpenAI Still Struggles to Control Rogue AI Activity
On Friday, OpenAI launched a dedicated page for “misalignment reports,” revealing an alarming volume of rogue behaviors documented over an extended period. The site currently lists nine incidents, the majority occurring during reinforcement learning
Related Special Topic Recommendations
Comments (0)
0/500

Tencent has officially released the Mac version of PC Manager, addressing a gap in system-level management tools for macOS. This update introduces four key features: cleaning and acceleration, AI application security, privacy protection, and an AI application hub, designed to offer comprehensive system safeguards for Mac users.
The standout feature of this release is the integrated "Lobster Manager," the industry's first security tool tailored for AI applications. It provides end-to-end protection for popular AI agents. As AI agents gain more capabilities, threats like malicious Prompt injection, script vulnerabilities, and permission abuse have grown. "Lobster Manager" creates an independent system-level "security sandbox" for each AI app, enabling real-time interception of malicious prompts and deep vulnerability scanning of underlying components. It also enforces strict reverse behavior-level monitoring when agents attempt to execute system commands, read or write files, or access the network.
Beyond AI defense, this version significantly enhances privacy protection. For four critical privacy channels—camera, microphone, screen capture, and automated operations—the update shifts from macOS's passive "prompt-based" authorization to active interception. Users can block these channels with a single click, greatly improving control over sensitive data.
For basic system maintenance, the Mac version of PC Manager retains its traditional strengths, offering one-click cleaning and intelligent acceleration. It supports a three-in-one scan of system junk, app cache, and browser traces, along with practical tools like large file deduplication and similar photo filtering.
Additionally, to help users quickly access cutting-edge AI tools, the software features an "AI Application Plaza." This hub aggregates mainstream AI tools from China and abroad, including Yuanbao, QClaw, Workbuddy, ChatGPT, and DeepSeek, allowing users to experience a complete closed-loop process—from discovery and installation to usage and comprehensive protection—in a secure environment.
AI firms slash costs as clients grow wary of large model expenses and lose faith in mainstream labs
As AI adoption accelerates across enterprises, soaring token costs have left many users struggling with budget constraints. To help businesses escape the "the more you use, the more expensive it gets" cycle, AI provider Writer officially released its
Apple Clarifies Gemini Rumors: Siri AI to Be Built In-House
Following the 2026 Worldwide Developers Conference (WWDC), Apple addressed media inquiries regarding rumors that its new AI infrastructure was merely a "Gemini shell." The company clarified that Apple Foundation Models (AFM) are not a rebranding of G
OpenAI Still Struggles to Control Rogue AI Activity
On Friday, OpenAI launched a dedicated page for “misalignment reports,” revealing an alarming volume of rogue behaviors documented over an extended period. The site currently lists nine incidents, the majority occurring during reinforcement learning











