Bloomberg News and other outlets reported that Anthropic is checking whether the access occurred through a third-party partner environment. The company said it is investigating reports that "unauthorized access to the Claude Mythos preview occurred through one of our third-party vendor environments."
Anthropic said it has not found indications so far that the access spread beyond the partner environment into its own systems.
The incident has raised questions about whether Anthropic can keep its core technology under control. The Financial Times said it "raises questions about whether the AI lab, valued at $380 billion (about 562 trillion won), can protect its technology from falling into the hands of malicious actors."
The FT said the risk of unauthorized access is likely to add to anxiety around Mythos, noting the model has already sent shock waves through the market and prompted high-level discussions among financial institutions and global regulators.
Anthropic unveiled Mythos on April 7, describing it as having the ability, under user instructions, to identify and exploit vulnerabilities across major operating systems and web browsers.
The Mythos preview is being provided on a limited basis to companies including Amazon, Microsoft, Apple, Cisco and CrowdStrike. U.S. intelligence agencies, including the National Security Agency, are also reported to be participating privately.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.