As AI-driven security risks grow and concerns spread that advanced AI models such as Anthropic’s Mythos could be used for cyberattacks, some expect security companies’ growth potential to increase.

The Information recently reported that CrowdStrike and Palo Alto Networks shares have each risen by about 20 percent over the past month. BlackRock’s IGV index, which tracks the software sector, rose by about 10 percent over the same period.

Despite the gains, the stocks’ valuation multiples are at the same level as a year earlier. CrowdStrike trades at about 19 times next year’s forecast revenue, while Palo Alto Networks is at around 11 times. The Information reported that this shows investors still do not properly value the unique growth potential of cybersecurity companies.

Anthropic is not releasing Mythos to the general public over concerns it could be used for cyberattacks, and is providing it for limited purposes only to some government agencies and companies. Under the name Project Glasswing, Anthropic is working with multiple partners to use Mythos, focusing on strengthening software security.

Security companies such as CrowdStrike, Palo Alto Networks and Zscaler were also included on the list of companies that can access Mythos.

The Information reported that these companies are in a favorable position because they can be prepared to meet rising security demand if Mythos is released more broadly.

Eric Heath (에릭 히스), a cybersecurity analyst at KeyBanc Capital Markets, said corporate cybersecurity spending has been relatively weak over the past 2 years. He said companies will increase cybersecurity spending because of the risks posed by Mythos, and added that the security spending environment will become much better than before.

There are concerns about competition from AI model companies, but many analyses still say existing security companies have the advantage so far.

Citizens Bank analyst Rustam Kanga (루스탐 캉가) said Anthropic and OpenAI models can currently only identify threats and cannot respond, while existing security companies have sensor software already installed on corporate devices and large volumes of proprietary data collected through it, making them difficult for AI model companies to replace.