Anthropic's leaked AI coding tool has been cloned over 8,000 times on GitHub despite mass takedowns

2026-04-01 · Source: The Decoder · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Fundamental Awareness, quick

Summary

Anthropic has initiated over 8,000 GitHub takedowns of its leaked AI coding tool's source code, according to the Wall Street Journal. The accidental leak, which included the "harness" and a "dreaming" function for task consolidation, allowed competitors to access techniques used to control Anthropic's AI models as coding agents. Despite the takedowns, one programmer successfully rewrote the code in different languages using AI tools, ensuring its continued availability. This incident highlights the rapid and uncontrollable spread of leaked code in the AI era, posing significant challenges for containment. The leak's timing is particularly problematic for Anthropic, which is planning an IPO at a $380 billion valuation, and follows closely after another human error-induced leak concerning its new AI model, Mythos.

Key takeaway

For CTOs and VPs of Engineering evaluating IP protection strategies, this incident underscores the extreme difficulty of containing code leaks in the age of AI. Your security protocols must extend beyond initial prevention to include rapid response and legal action, acknowledging that complete eradication of leaked code is unlikely. Prioritize securing content management systems and developer environments to mitigate human error, which remains a primary vector for such breaches.

Key insights

AI code leaks spread rapidly and are nearly impossible to fully contain once public.

Principles

• Code leaks are amplified by AI tools.
• Human error remains a critical vulnerability.

In practice

• Implement robust content management security.
• Prepare for rapid code dissemination post-leak.

Topics

• Anthropic
• AI Coding Tool
• Source Code Leak
• GitHub Takedowns
• AI Agents

Best for: CTO, VP of Engineering/Data, Executive, Tech Journalist, Investor, Director of AI/ML

Related on AIssential

• Anthropic Accidentally Leaked 512,000 Lines of Claude Code — Accidental source code leaks can rapidly spread and become irreversible, even with internal preventative measures.
• Mythos AI: Anthropic Investigates Unauthorised Access Claims — Anthropic's powerful Mythos AI, capable of chaining software vulnerabilities, faces unauthorized access claims, raising national security concerns.
• I Read Every Line of Anthropic’s Leaked Source Code So You Don’t Have To. — A configuration oversight led to the public exposure of Anthropic's proprietary Claude Code codebase.
• Anthropic's secret 'Mythos' model — Anthropic's 'Claude Mythos' AI, with advanced cyber capabilities, was accidentally leaked, signaling a new frontier model tier.
• Anthropic invites 150 more organizations into Project Glasswing — Claude Mythos is an AI model capable of advanced software vulnerability detection and exploitation, raising significant security concerns and regulatory attention.
• Anthropic took down thousands of GitHub repos trying to yank its leaked source code — a move the company says was an accident — Accidental source code leaks can trigger widespread, unintended consequences through automated takedown processes.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.