AI-Powered Malware: Challenges and Protection Strategies

Google’s Threat Intelligence Group identified experimental malware called PROMPTFLUX. This malware could potentially ask Gemini to rewrite its own code. One version of PROMPTFLUX was designed to do this every hour. The purpose of this continuous rewriting is clear: it makes the malware harder to detect.

Understanding Malware Rewriting

Security software typically looks for known patterns in malicious code. If that code continuously changes, the malware becomes a moving target. Although this doesn’t make it invisible to current tools, it poses a challenge for some detection methods.

An important point to note is that when Google discovered PROMPTFLUX, it was still in development. Researchers had not yet seen successful compromises of any device or network, and Google took steps to disable assets related to the activity.

AI’s Role in Malware

Following this discovery, Google documented instances of AI being used in malware during live attacks. It also identified an Android backdoor capable of using AI to understand the phone’s activity and assist in decision-making.

AI-powered malware like PROMPTFLUX exemplifies how attackers can use AI to continuously alter malicious code, making detection more difficult.

How PROMPTFLUX Operated

PROMPTFLUX was a VBScript-based project uncovered by Google in June 2025. A notable component was the “Thinking Robot,” which could contact Gemini for new obfuscation techniques, intended to make its code more elusive to security software.

Several variations of PROMPTFLUX were found, including one that instructed Gemini to rewrite the malware’s entire source code hourly while maintaining its operational parts. This creates a moving target for security software, complicating detection.

Antivirus vs. AI Malware

It’s crucial to realize that AI malware doesn’t render antivirus protection obsolete. Signature detection, behavioral analysis, and real-time monitoring remain vital in cybersecurity. Security software like Microsoft Defender employs these methods, along with machine learning, to identify new threats even if they don’t match a known signature.

Changing the code may not completely evade detection. Suspicious behavior can still alert security tools once malware exhibits dangerous activity.

PROMPTSTEAL in Live Attacks

While PROMPTFLUX was experimental, PROMPTSTEAL marked a new development. Google identified a Russian group deploying PROMPTSTEAL against targets. Unlike PROMPTFLUX, PROMPTSTEAL queries AI for Windows commands that it can execute, demonstrating a shift in malware interaction.

PROMPTSPY and Mobile Devices

PROMPTSPY represents advanced AI use in malware, directed at Android devices. It employs AI to interpret screen activity and decide how to interact with it. Google confirms it has mitigated the actors and prevented PROMPTSPY from appearing in Google Play apps.

Automation and Future Threats

Google’s reports highlight the trend towards AI-enabled automation in attacks. Example cases show attackers using AI to automate processes like vulnerability scanning and credential harvesting. However, fully autonomous AI attacks have not yet been observed.

Despite this, the role of AI in facilitating these tasks is growing, presenting a challenge for security teams.

Strategies for Protection

Here are actionable tips to protect against evolving threats:

• Use antivirus protection: Look for software with real-time protection and behavioral detection.
• Keep protections enabled: Ensure real-time and cloud protection are active.
• Enable automatic updates: Keep operating systems and applications up-to-date.
• Avoid executing random commands: Never run commands simply because a website suggests it.
• Heed security notices: Don’t ignore warnings from browsers or systems when downloading files.
• Select secure apps: Use trusted sources for apps and browser extensions.
• Use password managers: Implement multifactor authentication and passkeys.
• Maintain backups: Regularly backup important files separately.
• React to infections: Disconnect infected devices and conduct security scans promptly.

Key Takeaways

AI-powered malware introduces complex challenges in cybersecurity. While antivirus software continues to adapt, proactive measures can enhance your protection. As the technology progresses rapidly, it’s important to remain vigilant and be prepared to respond to potential threats.