Anthropic releases limited version of its AI model Mythos

Limited Access for the Public

Experts view Mythos as a model with extensive capabilities in the field of cybersecurity. According to experts, the AI model can identify and exploit security vulnerabilities in software. The AI application can, according to experts, identify and exploit security vulnerabilities with previously unattainable speed and precision. From the experts' perspective, the system is therefore suitable for both defensive security work and offensive attacks.

German constitutional protectors and security policymakers had warned that criminals or state actors could use Mythos for cyberattacks, for example on banks or energy facilities. The warning referred to the fundamental capabilities of the model line. With the limited version Fable 5, Anthropic appears to be attempting to find a middle ground between broader availability and risk limitation.

Warnings of Cyberattacks

The release comes against the backdrop of a broader debate about how to handle powerful AI systems. In recent months, security agencies and research institutions have repeatedly pointed to risks associated with modules for the autonomous discovery and exploitation of software vulnerabilities. Anthropic itself had already emphasized in April that the full model line must be handled with particular responsibility.

Observers view the step as a provisional compromise: on the one hand, the technology is to be made accessible to a limited group of users; on the other, its capabilities are to be reduced compared to the originally announced version. Whether these limitations will suffice to address the concerns of security agencies and political decision-makers remains to be seen.

Open Questions About Further Development

The debate surrounding Mythos is part of a series of similar discussions that have recently affected other manufacturers as well. At its core, the question is how much autonomy an AI system should have in identifying and exploiting security vulnerabilities, and what control mechanisms are required to prevent misuse.

Anthropic announced that it will continue to monitor developments and gradually adjust the scope of features of Fable 5. Further information regarding future availability, pricing models, and planned partnerships was not provided in the announcement. A statement addressing the specific concerns of German security policymakers has also not yet been issued.