AI accelerates exploitation of PHP2Shell vulnerability

Cybersecurity teams and web administrators are racing to update PHP systems after the publication of several security flaws. Attackers began attempting to exploit these weaknesses within hours of the disclosure to breach internet-connected systems. The vulnerability, known as PHP2Shell, gained attention because artificial intelligence helped researchers find a practical method to exploit it. While the underlying issues had been known for years, they were previously considered too complex to be used directly in an attack. Cloudflare analysis detected millions of exploitation attempts within hours of the vulnerability information being published. This situation illustrates how AI has the potential to accelerate attacker efforts in exploiting software flaws. Despite the sharp increase in attack attempts, this does not necessarily mean a mass hacking event has occurred. The PHP Foundation stated that it has not found any successful exploitation. Several public examples of exploits were also deemed ineffective. Researchers at the security company Calif identified PHP2Shell and developed a way to exploit it in early August.

They used an AI model through the OpenAI program Patch the Planet, but the process still required human expertise to understand the code and ensure the weakness could be utilized. Some of the problems underlying this flaw were actually identified in 2024. At that time, researchers had not found a practical way to exploit it. The capability of AI helped open the path to utilizing a weakness that was previously difficult to turn into a real attack. Researchers at OpenAI and the security company Wiz also independently found related vulnerabilities. According to Wiz security researcher Sagi Tzadik, the PHP team had known about this weakness for years but had not addressed it because a practical exploitation method had not been found. AI developments make vulnerability analysis faster and increase pressure on developers and administrators with limited resources. Microsoft security updates in October 2026 recorded a record 1,453 vulnerabilities. Bank of America also recruited 1,000 consultants to fix weaknesses found by Anthropic's AI Mythos. However, AI does not completely change the basic principles of cybersecurity. The technology accelerates the search and analysis of weaknesses, but complex research still requires human expertise.