February 25, 2026 Israeli software development company JFrog’s shares fell 24.94 per cent on Friday after Anthropic introduced Claude Code Security, a new feature designed to identify software vulnerabilities and recommend patches. The launch triggered broader selling across cybersecurity stocks, reflecting investor concern that AI-native tools could compete with established security platforms.
Anthropic said Claude Code Security can detect vulnerabilities, validate findings and suggest fixes, with recommended patches reviewed and approved through human processes rather than executed autonomously. The announcement weighed on several security names, including SentinelOne, which fell 4.35 per cent, and Varonis, down 7.68 per cent.
Jefferies analyst Joseph Gallo wrote that he continues to view cybersecurity as a likely beneficiary of the AI era, but estimated that “headwinds will increase before there is clarity regarding the impact of AI on cybersecurity.”
JFrog, led by cofounder and CEO Shlomo Ben Haim, provides software development and DevOps solutions. The stock had been one of the strongest Israeli performers on Wall Street in 2025, rising 112 per cent that year. However, even before Friday’s decline, shares were down 19.5 per cent in 2026. Following the latest drop, the stock has lost 39.6 per cent since the start of the year, reducing its market capitalization to about $4.5 billion, a decline of nearly $3 billion.
Raymond James told Yahoo Finance that the selloff appears exaggerated and characterized the risk as short term, maintaining a positive view of the company based on its platform resilience and strategic positioning.
Anthropic’s influence on markets has also been visible this week. IBM shares fell 13 per cent on Monday after the company highlighted how its Claude Code tools can accelerate refactoring of applications written in COBOL, a language still widely used in government, airline and financial systems.
In a blog post, Anthropic noted that COBOL applications remain prevalent, that developers proficient in the language are difficult to find, and that migration efforts are often costly and complex. “AI can assess which components are safe to move and which need careful handling. Areas with accumulated technical debt get documented before they become migration surprises,” the company wrote.
The premise is not new. IBM previously introduced watsonx Code Assistant for Z to help convert COBOL to Java, and major cloud providers including AWS and Microsoft, along with companies such as Kyndryl and NTT, have reported initiatives aimed at modernizing mainframe workloads.
