Korean Giants Secure Anthropic's Mythos Access

Samsung Electronics, SK Hynix, and SK Telecom are set to become the first Korean companies to join Anthropic's closed cybersecurity program, 'Project Glasswing.' On June 2, Boannews reported exclusively that the three firms secured access to 'Mythos,' Anthropic's top-tier AI model, through the program.

Often dubbed an 'AI hacker,' Mythos finds software vulnerabilities faster than human experts. Because of that risk, Anthropic had opened access to only about 50 major tech and financial firms worldwide, and this is the first time Korean companies have made the list. Korea's Ministry of Science and ICT had also been pushing for entry through its AI Strategy Committee.

Samsung and SK Hynix Bet on Chips, SKT on Equity

The three firms join for two different reasons. For Samsung and SK Hynix, it comes down to semiconductor security. As AI demand explodes, vulnerabilities in chip design and production software have become a matter that can make or break a national industry. With the two firms' combined market value exceeding half of the KOSPI, a production halt could ripple into industry-wide paralysis. By using Mythos, which flagged a flaw undetected for 27 years, they aim to audit their own software before attackers do.

SK Telecom's case is different. Its roughly 0.3% stake as an early Anthropic backer paved the way into Project Glasswing. Analysts value that stake at up to 4 trillion won. Yet with no concrete security project underway, turning that ticket into real business remains unfinished work.

Washington's Gatekeeping and Korea's 'AI Security Sovereignty'

The entry is not all smooth sailing. The White House, viewing Mythos's power as tied to national security, has been restraining any expansion of access. Anthropic reportedly proposed widening access from about 50 institutions to roughly 70, but the White House blocked it, and outside the U.S. the UK's AI Safety Institute is effectively the only body with formal access. As covered in our earlier report, even the EU failed to clear the threshold.

That is why Korea's Ministry of Science and ICT is pursuing talks with Anthropic while also preparing a homegrown AI security ecosystem. The worry is that if access to core models concentrates in one nation, technological dependence deepens. The question now is whether this trio's Project Glasswing entry will mark a starting point for Korea's 'AI security sovereignty' debate.