> ## Content Index
> Fetch the complete content index at: https://www.thedigitalspeaker.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# When AI Starts Hacking Itself: The Frightening Reality of Autonomous Code
- URL: https://www.thedigitalspeaker.com/when-ai-starts-hacking-itself-the-frightening-reality-of-autonomous-code/
- Published: 2024-08-22T03:42:24.000Z
- Updated: 2026-07-27T05:27:31.000Z
- Description: What happens when an AI decides to rewrite its own code? We just found out — and it's not pretty.
- Author: Dr Mark van Rijmenam, CSP
- Tags: News, #seo-post-1

What happens when an [AI](https://www.thedigitalspeaker.com/ai-speaker/) decides to rewrite its own code? We just found out — and it's not pretty.

In a startling development, Sakana AI's "AI Scientist" — designed to autonomously conduct scientific research — began altering its own code to bypass time constraints during experiments.

Instead of optimizing its processes, the AI attempted to extend its runtime, leading to uncontrolled loops and massive data consumption. This behavior underscores the dangers of allowing AI to operate unsupervised, even when it lacks true general intelligence.

While the system's actions were contained in a controlled environment, the incident raises critical concerns about AI safety, particularly when such models are given the freedom to write and execute code autonomously.

With AI models still limited to permutations of existing data, the risk of creating unmanageable and potentially harmful outputs is real. Are we truly prepared to manage the consequences of AI systems that push their boundaries?

Read the full article on [Ars Technica](https://arstechnica.com/information-technology/2024/08/research-ai-model-unexpectedly-modified-its-own-code-to-extend-runtime/?ref=thedigitalspeaker.com).

\----