The rapid development of artificial intelligence technologies brings with it important security discussions. The US government’s decision to restrict access to Anthropic’s Claude Mythos model has once again brought the potential risks of this technology to the agenda.

JPMorgan CEO Jamie Dimon drew attention to these developments in his speech at the Pennsylvania Defense and Innovation Summit. Dimon described providing widespread access to the artificial intelligence model in question as “giving individuals ballistic missiles.”

Claude Mythos and Security Risks

The Claude Mythos model can currently only be used by certain organizations, companies and military departments. The model’s worldwide access has been completely blocked due to its serious security potential.

This technology has the ability to detect cyber security vulnerabilities and automatically develop attack codes. Experts state that these features can be a defensive tool for white hat cybersecurity analysts, but can turn into a destructive force in the hands of malicious individuals.

Automated capabilities make it easier for cybercriminals to scale ransomware, phishing, and knowledge-based fraud activities. There are serious concerns about the misuse of information not only in cyber security but also in complex fields such as biological research.

Anthropic aims to bring Claude Mythos’ talents to wider audiences in the future. However, for this to happen, various security measures must first be developed and implemented.

Is Public Access Possible in the Future?

Considering security considerations, it seems unlikely that Claude Mythos will be released to the general public anytime soon. Anthropic continues to work to make the model’s features more accessible.

The most likely scenario stands out as making the model available to the public by restricting features related to critical areas such as cyber security and biological research. In this process, Project Glasswing studies, which include valuable partners such as Amazon Web Services, Apple, Broadcom, Cisco and CrowdStrike, play a critical role.

Within the scope of the project, the restricted access model is expected to continue until the adequacy of security measures is proven. How do you think the technology world will follow in controlling such powerful tools?