Leaks about threats to Anthropic's new AI model have brought down cybersecurity stocks

Shares of cybersecurity companies fell in trading on March 27 - after Fortune magazine published an article stating that a new AI model currently being tested by Anthropic could be used by hackers to bypass existing cyber defenses, Bloomberg writes.

Against this background, the securities of CrowdStrike at the intraday low collapsed by 7.8% (at the time of publication they are down 6.3%), shares of Palo Alto at the moment were losing 8.2% (then their decline slowed to 6.6%), Zscaler was losing 9.5%, and the securities of Cloudflare at the low fell by 4.78%. Overall, the Global X Cybersecurity Sector ETF ETF was down 6.1% on March 27, bringing its year-to-date decline to more than 20%, Bloomberg points out.

The Fortune article that brought down the papers of cybersecurity companies quoted a draft of a post from Anthropic's blog that said, Anthropic believes the new AI model the company is now testing "poses unprecedented cybersecurity risks." Fortune explained that it gained access to this draft through an unsecured and publicly available database. An Anthropic spokesperson confirmed that the company is testing a new AI model, and also admitted that a draft of the blog post was indeed made available for public viewing due to a bug.

The Fortune article suggests that the AI model in question is Anthropic's AI model - dubbed Claude Capybara - which is currently being tested by a small group of users who have gained early access to it. Given the capabilities of the new AI model, Anthropic plans to share the results of its testing with cybersecurity companies to help them improve their defenses before its release, the company said in a blog post.

Amid the sell-off in cybersecurity securities, Wall Street analysts rushed to support companies from this sphere, Bloomberg notes.

"We believe Anthropic is also seeking to limit the use of its product by hackers. This is basic security practice and the minimum expected standard for such a product," the agency quoted Bernstein analyst Peter Weed as saying.

The market is "misinterpreting the news" about the security of the new AI model, Todd Weller of Stephens also noted. He called it a buying opportunity. Equity Armor Investments portfolio manager Joe Taigai, whose funds invest in Palo Alto, CrowdStrike and Fortinet, shares this view: "The only thing that doesn't add up in this logic [of the market selling off on the news of the new AI model and its security capabilities]: if this is really a cybersecurity threat, doesn't it make sense that the best companies in the field would fight it?" he wondered. - he wondered.

Such a sell-off in the cybersecurity market is not the first, CNBC reminds. The sector has already fallen victim to concerns about the destructive impact of artificial intelligence in February. Back then, the securities of companies in the field fell after Anthropic unveiled a new code-scanning tool built into Claude's AI model. A broader segment of the software market has also come under pressure from these innovations, CNBC notes.

This article was AI-translated and verified by a human editor