A patch-comparison sheet traces the teeth of a key, illustrating how published repairs can guide exploit development.
Image by CryptoSlate

Google uncovers AI hacking tool that turns security fixes into cyberattacks

Researchers discovered an attacker-controlled system capable of identifying software vulnerabilities, generating exploits and executing attacks automatically.

Quick Take

  1. Google uncovered an AI hacking framework capable of turning software security patches into exploits and launching attacks automatically.
  2. The system could help hackers target vulnerable websites before security updates are installed, sharply reducing the time available for protection.
  3. Google warns that increasingly autonomous hacking tools could accelerate cyberattacks as criminals automate vulnerability discovery and exploitati

Google has uncovered an AI-powered hacking system capable of transforming software security fixes into automated attacks against vulnerable websites.

In a Nov. 16 report, Google Threat Intelligence Group and Mandiant revealed that attackers were using AI frameworks to identify vulnerabilities, generate malicious code, and execute exploits with minimal human involvement.

The researchers discovered one such system, dubbed exploit_agent, on an exposed staging server containing multiple AI models connected to cybersecurity research and exploitation tools.

The framework was configured to monitor software updates, compare patched code with older versions, and identify weaknesses that developers had recently repaired. It could subsequently generate attack code and execute it automatically, allowing attackers to target websites that had not yet installed the relevant security updates.

This approach exploits the information developers publish when addressing software vulnerabilities. Changes made to repair a flaw can inadvertently reveal how the underlying weakness could be exploited.

Google said the technique lets attackers focus on known vulnerabilities rather than spend time discovering previously unknown security flaws.

The company's investigation also uncovered stolen victim data and malicious scripts capable of conducting internal network reconnaissance, brute-force attacks and installing webshells that provide remote access to compromised servers.

Some tools could remove webshells left by competing attackers, potentially helping the operator retain control over compromised systems.

Google found evidence that the operator was particularly interested in French websites associated with media, entertainment, and education. One recovered example showed the framework examining access-control protections in Spectra, a WordPress plugin used for website development.

From patch to exploit

Google describes exploit_agent comparing patched code, identifying potential weaknesses, drafting exploits and executing them automatically.

A published fix can guide attack development while website owners are still deploying the update.

The diagram shows Google's configured pipeline, not measured timing, a success rate or proof of an entirely autonomous campaign.

The discovery marked the first time Google had identified attackers using AI frameworks to fully automate vulnerability discovery and exploitation without human intervention throughout the process.

North Korean hackers turn AI development tools against crypto companies

Google's report also revealed a separate campaign in which suspected North Korean hackers exploited AI-powered development tools to infiltrate cryptocurrency organizations in North America and Europe.

The attackers, tracked as MIDNIGHT NEPTUNE, combined AI-generated identities with fraudulent recruitment campaigns to gain access to corporate systems.

They created fake recruiter profiles on professional networking platforms using AI-generated photographs, then directed prospective victims to fraudulent meeting websites.

Those websites instructed victims to run supposed troubleshooting commands that instead executed malicious scripts.

Once inside an organization, the attackers accessed internal code repositories and modified configuration files used by Anthropic's Claude Code development assistant.

The altered files were designed to automatically download a backdoor known as SOMBERMEME whenever employees launched Claude Code.

Google also discovered that the attackers disguised GitHub access tokens using names associated with Microsoft Copilot, concealing their movements within compromised development environments.

Related Reading

Companies using AI report 120% more bugs with no detected repair slowdown, study finds

The campaign expanded on North Korea's established use of AI to generate fake resumes, prepare for job interviews, and impersonate Western technology workers seeking remote employment.

Google said the latest intrusions showed how attackers exploited AI development tools already embedded in corporate workflows to distribute malicious software.

Meanwhile, the emergence of autonomous exploitation frameworks threatens to accelerate attacks against organizations running outdated software.

Google warned that automating vulnerability analysis and exploit execution could substantially reduce the time between disclosing a security flaw and exploiting it, leaving website operators with increasingly narrow windows to install protective updates.