Skip to content

Google DeepMind Unleashes AI to Automatically Fix Code Vulnerabilities!

DeepMind's AI: Revolutionizing Code Security & Patching Vulnerabilities

Google DeepMind unveils AI agent that automatically repairs code vulnerabilities

Hello everyone. This is Jon. AI technology continues to evolve daily, and on October 6, 2025, Google DeepMind announced a new AI agent called "CodeMender." This tool automatically finds and fixes security weaknesses (vulnerabilities) hidden in software code. In the world of programming, vulnerabilities are a common source of attacks by hackers, so having an AI autonomously fix these vulnerabilities is groundbreaking. In this article, I'll explain the details of CodeMender in a way that's easy to understand even for beginners. Let's start with an overview.

Recommended for those who want to start automating with no coding!
With Make.com (formerly Integromat)...
📌 Integrate major tools like email, Slack, Google Sheets, and Notion all at once
📌 Automate complex tasks with just drag and drop
📌 A free plan is also available, so you can try it out for yourself.
If you're interested, here's the details:
What is Make.com (formerly Integromat)? How to Use It, Pricing, Reviews, and Latest Information [2025 Edition]

What is CodeMender? Explaining the basics

CodeMender is an AI-based agent developed by Google DeepMind that automatically analyzes software code and fixes security issues. The term "agent" here refers to an autonomous system that thinks and acts on its own. While traditional tools only detect vulnerabilities, CodeMender goes beyond that by rewriting the code to eradicate the issues.

The announcement was made on October 6, 2025, with details published on DeepMind's official blog. This AI is based on an advanced model called Gemini Deep Think. Gemini is a type of Google's Large Language Model (LLM) that is highly capable of logically solving complex problems. CodeMender leverages this to scan huge numbers of lines of code (sometimes over 4.5 million lines) and identify potential vulnerabilities.

For example, software vulnerabilities are security holes that arise due to coding errors or outdated design. Hackers can exploit these vulnerabilities, potentially resulting in data leaks or system outages. CodeMender uses AI to automatically create and apply patches (corrective programs) to address such issues. It also has the ability to safely rewrite the entire code to prevent similar vulnerabilities all at once.

The emergence of such AI tools reduces the burden on developers. Incidentally, there is also a similar AI tool called Gamma, which can instantly create documents and slides. If you want to use AI to streamline your creative work,Find out more about Gamma hereGive it a try. We are entering an era where AI can help us with specialized tasks, just like CodeMender.

Key Features of CodeMender

Here's a summary of CodeMender's features:

  • Auto-detect: Scans code to discover unknown vulnerabilities. Accuracy is improved through collaboration with DeepMind's previous projects (Big Sleep and OSS-Fuzz).
  • Patch Creation: Apply fixes to problem areas, suggesting fixes in a way that's easy for a human developer to review.
  • Code rewrite: We restructure the code to eliminate entire classes of vulnerabilities rather than just fixing them, for example, by addressing memory safety issues at their source.
  • Large-scale response: Works with codebases of millions of lines. Proven in open source projects.

These features combine AI reasoning capabilities with program analysis technology. It may sound difficult for beginners, but the idea is that AI acts like a "code doctor."

Latest results and actual use cases

While CodeMender is still in the research stage, it is already proving successful: As of the announcement on October 6, 2025, DeepMind had submitted 72 security fixes to open source projects, totaling more than 4.5 million lines of code, including critical software like SQLite.

Looking at the timeline, DeepMind has been working on vulnerability detection for some time. For example, in 2024, a case was reported in which AI discovered a zero-day vulnerability (an unknown security hole). CodeMender has evolved this to automate the process from discovery to correction. In news from October 7, 2025 onwards, media outlets such as SiliconANGLE and CSO Online reported on this progress, emphasizing the benefit of allowing developers to focus on their core business.

X (formerly Twitter) became a hot topic immediately after the announcement. Developers and AI enthusiasts responded positively, saying things like, "This will change the future of security," and "The era of AI fixing code has arrived." However, since X's posts are personal opinions, it's important to verify the facts based on official announcements.

Specific example

Examples of how CodeMender can be used:

  • Contributions to open source projects: 72 fixes merged and implemented in the production software.
  • Large code base: Automatically remediate vulnerabilities in repositories with over 450 million lines of code.
  • Preventative approach: Patch new vulnerabilities immediately and prevent future attacks.

This is laying the foundation for AI to play an active role in the field of cybersecurity. As of October 9, 2025, the latest information is that DeepMind is reportedly working on further improvements.

CodeMender's future prospects and points to note

CodeMender is expected to be a tool that can keep up with the pace of vulnerability discovery using AI. This type of defensive AI is important because AI itself can be used as a hacking tool. In the future, it may become something that companies and individual developers can use on a daily basis.

However, since it is currently in the research phase, there are no plans to release it to the public. Human review is required, and AI corrections are not always perfect. Continuous improvement is necessary to increase reliability.

If you want to use AI to streamline your document creation, why not give Gamma a try? Learn more about GammaFrom here.

Jon's Summary

CodeMender is a great step forward in using AI to automate code security. It has the potential to reduce the burden on developers and create a more secure software environment. However, it's important to keep up with technological advances and consider the balance between humans and AI. I'm looking forward to future updates.

Reference sources

Related posts

tag:

Leave a comment

There is no sure that your email address is published. ■ Required fields are marked