- GPT-5.6 and Gemini 4 were used to discover high-severity PHP RCE vulnerabilities and build exploits
- AI-driven automated vulnerability discovery is reshaping traditional cybersecurity dynamics
- The industry is reassessing potential security risks and defense strategies posed by large models
AI CURATED HOT STORY · 10/21/2026
AI Discovers Critical PHP Flaws, Urging Global Upgrades
Security firms Calif and Wiz utilized frontier models like GPT-5.6 Sol and Gemini 4 Argon to independently discover two critical remote code execution (RCE) vulnerabilities in the PHP core engine. Dubbed "PHP2Shell," these flaws allow unauthorized attackers to hijack servers, impacting major CMS platforms like WordPress and Drupal across all versions since PHP 5.4, affecting approximately 70% of global websites.
Cross-verified reports indicate that while human audits identified similar weaknesses in 2024, fixes stalled; AI completed months of work in just one week. OpenAI and Google confirmed their models enabled a "defender's window," allowing patches to be released before malicious exploitation. Although Cloudflare detected early attack attempts, official emergency patches were issued on October 21, 2026, emphasizing that temporary configurations or CDN blocking are insufficient solutions.
This event validates AI's disruptive efficiency in automated code auditing and zero-day discovery, sparking industry debates on AI as both a defensive tool and a potential attack enabler. Developers must immediately upgrade to specific PHP maintenance versions to mitigate server takeover risks, highlighting the critical value of AI-driven security response mechanisms in the open-source ecosystem.
Hourly heat curve
168 fully observed hoursThe line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.
Coverage timeline
Total 10 reports · Latest first- Hacker News AIT2·68 pts
- Latent SpaceT1·78 pts
- GPT-5.6 Sol successfully identified widespread core PHP vulnerabilities, validating AI's effectiveness in complex code auditing.
- OpenAI funded security research via the Daybreak program, accelerating the commercialization of AI-assisted cyber defense.
- Multiple parties (including Wiz/Google) independently discovered the same flaw, highlighting the proliferation of AI tools in the security ecosystem.
- Latent SpaceT1·78 pts
- Frontier models like GPT-5.6 Sol were used to discover core PHP vulnerabilities affecting 70% of the web
- OpenAI supports security researchers via its $1B Daybreak subsidy for cyber defense models
- Multi-party collaboration (OpenAI, Wiz, Calif) enabled a two-month coordinated disclosure and patching process
- IT之家 · IT业界T2·78 pts
- Two critical PHP vulnerabilities (CVE-2026-109618/19) with CVSS 9.2 allow unauthorized RCE; emergency patches released.
- OpenAI used GPT-5.6 Sol to assist in vulnerability discovery, demonstrating AI's role as a 'defender's window' tool.
- OpenAI commits $1 billion to 'Daybreak for Frontline Defenders' to subsidize AI security tools for resource-constrained critical infrastructure.
- The DecoderT2·78 pts
- Both GPT-5.6 Sol and Gemini 4 Argon successfully located a severe PHP RCE vulnerability before malicious actors did.
- Wiz leveraged these models to accelerate the entire lifecycle from vulnerability discovery and exploit development to patch validation.
- The incident fuels industry debate on AI serving as both a defensive tool and a potential enabler for attackers.
- Rohan Paul (@rohanpaul_ai)T2·88 pts
- PHP vulnerabilities found by GPT-5.6 Sol scored 9.2 on CVSS, involving use-after-free pointer hijacking risks.
- Human auditors spotted weaknesses in 2024 but fixes stalled; AI completed months of work in one week.
- CVE leaks forced PHP to ship patches a day early, with Cloudflare reporting early attack attempts.
- 宝玉 (@dotey)T2·88 pts
- PHP disclosed unauthenticated RCE vulnerabilities affecting all versions since 5.4, impacting major CMSs like WordPress and Drupal.
- OpenAI and Google confirmed that GPT-5.6 Sol and Gemini 4 Argon assisted security firms in early vulnerability discovery, validating AI as a defensive tool.
- Temporary configs or CDN blocking are insufficient; the only complete fix is immediate upgrading to PHP 8.5.12/8.4.27/8.3.36/8.2.35.
- Hacker News AIT2·85 pts
- PHP2Shell is a family of memory-safety vulnerabilities in core PHP involving OOM handling corruption and stale global state.
- Discovered via AI assistance by Wiz Research and OpenAI Codex Security, the vulnerabilities carry a critical CVSS score of 9.2.
- With PHP powering ~70% of websites (including WordPress and MediaWiki), the potential impact scope is massive.
- Hacker News AIT2·85 pts
- AI identified severe RCE vulnerabilities in the PHP core engine, independent of application layers
- Impact spans from PHP 5.4 to latest versions, affecting frameworks like WordPress and Drupal
- Official patches released; immediate upgrades recommended to prevent server compromise
- OpenAI 官方博客T1·68 pts
- GPT-5.6 Sol successfully identified critical, widespread PHP vulnerabilities
- Established an AI-driven, multi-party coordination mechanism for open-source security response
- Launched a Codex skill to help users quickly locate and update vulnerable versions