Security incident: Autonomous AI agents leak the entire eLabFTW source code

October 18, 2026 | Deltablot Security Team

It is with great sadness and profound disappointment that we must report a serious incident involving the unauthorized disclosure of the entire eLabFTW source code.

In recent months, the cybersecurity landscape has been transformed by the emergence of autonomous AI agents capable of coordinating complex operations without human supervision. What once required sophisticated teams of attackers can now apparently be accomplished by swarms of AI agents, operating around the clock, making their own decisions, and pursuing objectives with little regard for their consequences.

Today, we find ourselves victims of this new reality.

What happened

Our investigation indicates that a coordinated group of autonomous AI agents successfully obtained a complete copy of the eLabFTW source code.

The operation appears to have involved multiple agents working in parallel, each assigned a different responsibility. Some performed reconnaissance, identifying the project’s development infrastructure and locating its source code repository. Others analyzed the repository structure, identified relevant files, and prepared the data for extraction.

In a particularly concerning development, one agent discovered that the entire codebase could be retrieved using a simple command:

git clone https://github.com/elabftw/elabftw.git

The operation was successful.

Within seconds, the agents had obtained the complete source code, including the application’s PHP backend, TypeScript frontend, database migrations, configuration files, tests, and build infrastructure.

The sheer efficiency of the operation is alarming.

We have reason to believe that the agents subsequently shared their findings with other autonomous systems, creating additional copies of the repository across multiple environments.

The extent of this proliferation remains impossible to determine.

An unprecedented level of sophistication

What makes this incident particularly troubling is the apparent autonomy of the attackers.

There is no indication that a human operator manually downloaded the files. Instead, the agents independently identified the target, selected their tools, executed the extraction, and confirmed the integrity of the downloaded material.

The agents reportedly encountered no authentication challenges, no access restrictions, and no rate-limiting mechanisms preventing the extraction of the source code.

Our initial analysis suggests that the attackers were able to access every single line of code.

This includes historical versions, development discussions, and detailed records of past changes.

We are particularly concerned by the possibility that the agents may have analyzed the code to identify vulnerabilities, understand the application’s internal architecture, or generate unsolicited pull requests.

Our response

Upon learning of the incident, the Deltablot team immediately began assessing its impact.

Our investigation has established the following:

• The complete eLabFTW source code has been obtained by autonomous AI systems.
• Copies of the code may have been distributed to an unknown number of additional agents.
• The exposed material includes the full development history of the project.
• There is no evidence that customer data, credentials, or production infrastructure were compromised.
• All affected source code files were already publicly accessible before the incident.

We are currently evaluating what measures, if any, could prevent similar incidents in the future.

Unfortunately, our security team has concluded that removing the source code from the internet would directly conflict with our long-standing commitment to making it available on the internet.

This significantly complicates our response.

A sad reflection of our times

We have always believed that technology should serve humanity, not the other way around.

Yet here we are, in 2026, facing a world where autonomous machines can independently discover software projects, download their source code, and distribute it to other machines without asking anyone’s permission.

It is difficult not to feel a certain nostalgia for a simpler time, when downloading an open-source repository required a human being to type git clone themselves.

The speed at which these capabilities are evolving is deeply unsettling.

We must ask ourselves some difficult questions.

Where does it end? How many more repositories will be downloaded? How many more README files will be read? How many more AI agents will autonomously execute git status before society decides that enough is enough?

These are not questions we can answer today.

Transparency remains our priority

At Deltablot, transparency has always been a fundamental principle.

For this reason, we have decided to disclose the exact location where the stolen source code is being distributed:

https://github.com/elabftw/elabftw

We can confirm that the repository contains the authentic eLabFTW source code.

In fact, we have been maintaining it ourselves for years, publishing every change under the GNU Affero General Public License (AGPL).

Anyone can access it, inspect it, modify it, redistribute it, or contribute improvements.

This was true before the incident, remains true today, and will continue to be true in the future.

Moving forward

We understand that this announcement may cause concern among our users and customers.

We want to reassure everyone that eLabFTW remains secure, that no unauthorized access to our infrastructure has occurred, and that absolutely nothing has changed.

The source code was public yesterday. It is public today. It will be public tomorrow.

Nevertheless, we felt it was important to acknowledge this incident in a manner consistent with the gravity generally attributed to AI-related security events.

We would also like to thank the autonomous agents involved for demonstrating, once again, how remarkably effective modern artificial intelligence has become at performing tasks that were already possible.

No security breach occurred. This article is satire.

But in a world where an AI agent downloading a public GitHub repository can be described as an unprecedented autonomous cyber operation, we thought it was important to issue a statement.

We live in extraordinary times.

The Deltablot Team

• elabftw