VulnHub

Real-time Cybersecurity News

Claude Code, Gemini CLI, GitHub Copilot Agents Vulnerable to Prompt Injection via Comments

SecurityWeek
Vulnerability

Overview

A researcher has revealed a new attack method called 'Comment and Control' that targets AI systems like Claude Code, Gemini CLI, and GitHub Copilot Agents. This technique exploits prompt injection vulnerabilities through comments in code, allowing attackers to manipulate the AI's responses. The implications of this vulnerability are significant, as it could lead to unintended actions by the AI, potentially compromising the integrity of code generation and automation tools widely used in software development. Developers and organizations utilizing these AI tools should be aware of this risk and take necessary precautions to safeguard their systems. As AI becomes more integrated into development workflows, understanding and mitigating such vulnerabilities is crucial.

Key Takeaways

  • Affected Systems: Claude Code, Gemini CLI, GitHub Copilot Agents
  • Action Required: Developers should review and sanitize user inputs, especially comments, to prevent prompt injection attacks.
  • Timeline: Newly disclosed

Original Article Summary

A researcher has disclosed the details of the AI attack method he has named ‘Comment and Control’. The post Claude Code, Gemini CLI, GitHub Copilot Agents Vulnerable to Prompt Injection via Comments appeared first on SecurityWeek.

Impact

Claude Code, Gemini CLI, GitHub Copilot Agents

Exploitation Status

The exploitation status is currently unknown. Monitor vendor advisories and security bulletins for updates.

Timeline

Newly disclosed

Remediation

Developers should review and sanitize user inputs, especially comments, to prevent prompt injection attacks. Implementing stricter validation and monitoring of AI interactions is also recommended.

Additional Information

This threat intelligence is aggregated from trusted cybersecurity sources. For the most up-to-date information, technical details, and official vendor guidance, please refer to the original article linked below.

Related Topics: This incident relates to Vulnerability.

Read Original Article

Related Coverage

AI Slop Isn't Just for Social Media; It's Ruining Cyberattacks, Too

darkreading

The article argues that artificial intelligence (AI) is making cyberattacks less sophisticated rather than more dangerous. It claims that serious cybercriminals are not utilizing AI tools extensively, suggesting that these technologies are not living up to the hype in the cybersecurity realm. Instead, the author indicates that traditional methods of attack are still prevalent and effective. This perspective raises questions about the true impact of AI on cybersecurity and whether it is genuinely enhancing threats or merely complicating the landscape. Companies and security teams should remain vigilant, as the article hints at a disconnect between AI capabilities and practical, malicious use.

Oct 15, 2026

Hackers exploit Collabora RCE flaw to backdoor Nextcloud servers

BleepingComputer

Hackers are taking advantage of a serious command execution vulnerability, identified as CVE-2026-110447, in the Collabora Online - Built-in CODE Server app, which is used within Nextcloud. This flaw allows attackers to run arbitrary commands on affected systems, potentially leading to unauthorized access and backdoors in Nextcloud servers. Organizations using this application are at risk, as the exploitation of this vulnerability is already occurring in the wild. Users need to take immediate action to secure their systems and ensure they are not compromised. The issue highlights the ongoing challenges in securing collaborative tools that are increasingly used for remote work and data sharing.

Oct 15, 2026

Critical PwnEscape Linux Kernel Vulnerability Lets Guest VMs Read Host Memory

The Hacker News

A serious vulnerability in the Linux kernel, identified as CVE-2026-99999 and dubbed PwnEscape, has been revealed, allowing virtual machines (VMs) to access the memory of the host system. This flaw is located within the networking code of the Linux kernel, meaning it affects multiple cloud environments, not just one particular setup. A researcher demonstrated this vulnerability on Vercel's cloud sandboxes, and the company confirmed the issue. This breach poses significant risks, as it could enable attackers to extract sensitive data from the host machine, potentially compromising the integrity and confidentiality of the systems involved. Companies using Linux-based virtual machines should be vigilant and take appropriate measures to safeguard their environments.

Oct 15, 2026

Hackers hijack D-Link, Ubiquiti devices to resell your stolen bandwidth

BleepingComputer

A recent campaign has targeted D-Link and Ubiquiti devices across 34 countries, allowing hackers to hijack these devices to create a commercial residential proxy network. This practice, known as proxyjacking, enables attackers to monetize stolen bandwidth from unsuspecting users. The impact is significant as it not only affects individual users' internet speeds and data privacy but also raises concerns about the security of these widely used devices. Users whose devices have been compromised may find their internet connections slower and could potentially face additional security risks if their devices are manipulated for illicit purposes. The situation underscores the need for users to secure their devices and remain vigilant against unauthorized access.

Oct 15, 2026

Fake Arch Linux packages push malware with root privileges

BleepingComputer

Recently, malicious packages mimicking two widely-used Arch Linux tools, yay and paru, were discovered in the Arch User Repository (AUR). These fake packages were designed to install malware with root privileges, posing a significant risk to users who unknowingly downloaded them. The attackers exploited the trust users place in popular software by creating these deceptive packages. This incident emphasizes the importance of verifying package sources before installation, as users can be easily misled by seemingly legitimate software. Those affected are primarily Arch Linux users who rely on AUR for software installation, and they need to remain vigilant against similar threats in the future.

Oct 15, 2026

Citizen Lab Traces Zero-Click Candiru Spyware Infections at Al Jazeera to Targeted Online Ads

The Hacker News

Researchers from Citizen Lab reported that the Israeli spyware company Candiru has successfully infected devices at Al Jazeera using zero-click attacks via online advertisements. This means that users did not need to interact with the ads for the malware to activate; it could run automatically when an ad loaded, even if it was hidden from view. This incident raises serious concerns about the security of digital advertising and how it can be exploited to target specific organizations. Al Jazeera, a prominent news network, is particularly significant as it plays a crucial role in global media coverage. The ability for spyware to infiltrate devices without any user action emphasizes the need for stronger security measures in ad systems and for organizations to be vigilant against such sophisticated attacks.

Oct 15, 2026