Is the Internet security sector facing another disturbance? Anthropic will launch the Mythos model to the UK banking industry next week.

The Anthropic plan to release its highly anticipated Mythos artificial intelligence model to UK financial institutions next week. It is reported that the model has attracted high levels of attention from global regulatory bodies, and disclosures around its early testing capabilities have caused temporary market volatility, with investors reevaluating related risks. Previously, disclosures related to early testing versions of the model (such as "Claude Capybara") had sparked significant market reactions. In late March, as Anthropic gradually disclosed potential risks of the model, concerns arose in the market about the AI's ability to bypass existing network defense systems and achieve autonomous vulnerability discovery, leading to pressure in the cybersecurity sector and notable declines in stocks including CrowdStrike, Palo Alto Networks, and Zscaler. However, as more technical details were disclosed and the "Glasswing" project progressed, market sentiment began to correct. Investors gradually realized that AI not only could reshape attack methods but also strengthen the demand for cybersecurity. Supported by related partnership announcements, the cybersecurity sector experienced a rebound at one point, indicating a shift in market pricing from the early "single impact" phase to a reevaluation stage of "risk and opportunity coexisting." Realizing that the AI model is a powerful tool for discovering and potentially exploiting cybersecurity vulnerabilities, Anthropic is gradually expanding the "Glasswing project," which aims to grant specific organizations early access to the AI model. "This refers to the recent, next week," said Pipp Whit, head of Anthropic in the UK, Ireland, and Northern Europe in an interview. "As expected, my communication with UK CEOs has been very frequent over the past week." Earlier this month, Anthropic stated that it would restrict the release of the Mythos AI model to allow cybersecurity professionals the opportunity to test their defense systems using the model. According to Anthropic, thousands of "zero-day" vulnerabilities have been discovered during testing of Mythos, covering all mainstream operating systems and web browsers. Some of the initial participants in the Glasswing plan include large tech companies such as Amazon, Apple, Microsoft, and Cisco. Due to limited access and understanding of the model's capabilities, banks and government institutions have expressed concern. Shortly after the release of the Glasswing plan, US Treasury Secretary Scott Bennett and Federal Reserve Chairman Jerome Powell convened Wall Street executives for an emergency discussion on Anthropic's Mythos and similar AI models. It was reported that Citigroup, Morgan Stanley, Bank of America, Wells Fargo, and Goldman Sachs were all present. The meeting was not publicly announced beforehand and was hastily arranged, with regulators internally qualifying it as a "clear sign of impending systemic financial risk." On Tuesday, Bank of England Governor Andrew Bailey stated that global regulatory bodies need to swiftly assess the threats posed by the Anthropic model. The Bank of Canada has also convened banks and financial institutions to discuss related risks. Whit stated, "We have put our own security measures and restrictions in place for this product because we know how powerful it is." The UK Artificial Intelligence Security Institute announced on Wednesday that they have been granted permission to use the model. This government-supported agency responsible for assessing AI risks found that the model surpassed previous cutting-edge models in simulating multi-step network attacks. Whit also mentioned that Anthropic plans to open an office in London in the first quarter of next year. The company currently has around 200 employees in London and the new office will accommodate up to 800 employees.