Artificial intelligence (AI) corporations Anthropic said on the 2nd (local time) that it has expanded eligibility for Project Glasswing, a cybersecurity cooperation plan based on Mithos, to about 150 new institutions in 15 countries.

Mithos is an AI model with expert-level capabilities to detect security vulnerabilities. Project Glasswing is a security consortium that Anthropic formed when it disclosed in Apr. that it was developing the Mithos model.

Anthropic logo. /Courtesy of Yonhap News

Anthropic explained that the newly participating institutions span new industries such as power, water, healthcare, telecommunications, and hardware, which were lacking among the initial participants. However, these institutions must meet security requirements before being granted access.

Anthropic emphasized that while the countries of affiliation for the newly added partner institutions number 15, if these institutions encounter problems, more than 100 million people could be affected across borders, with potential repercussions for global security and national security.

Among the corporations joining as partners this time are believed to be domestic corporations and institutions, including Samsung Electronics and SK hynix, which recently joined as Anthropic investors.

Anthropic said that security checks using the Mithos model at about 50 initial partner institutions uncovered more than 10,000 security flaws rated "high" or "critical" in just a few weeks.

Anthropic is rushing this expansion because AI models are advancing faster than expected. Anthropic noted, "Within 6–12 months, several other AI corporations will also have models on par with Mithos," and warned, "They may release those models without safeguards to prevent misuse."

It added, "In such an environment, cyberattacks could occur far more frequently and in unpredictable forms," and pointed out, "Security personnel must prepare countermeasures to keep pace."

Fearing that Mithos could be misused by malicious hackers, Anthropic provided it in advance to vetted institutions such as major technology corporations, financial institutions, and governments so they can find and defend against software vulnerabilities.

※ This article has been translated by AI. Share your feedback here.