Oct 22·AI
Follow-up 3d ago7
94%
Critical PHP vulnerabilities expose 70% of web; patching underway globally
Two critical vulnerabilities in PHP, which powers 70% of websites globally, were publicly disclosed on October 23, 2026, sparking worldwide emergency patching efforts and temporary website shutdowns. Despite millions of exploitation attempts detected within hours, security firms reported no confirmed successful attacks as of late Thursday. The discovery, made using AI-assisted analysis, prompted governments and organizations to either apply patches urgently or take systems offline to prevent compromise.
Quick Facts
- Two critical vulnerabilities discovered in PHP programming language
- Vulnerabilities enable remote code execution and unauthenticated access
- Flaws affect approximately 70% of websites worldwide
- Patches released by PHP Foundation on Wednesday
- Global government and private sector websites temporarily taken offline
Related Insights
Cyber Security Agency urges organisations to patch critical PHP flaws immediately CNA · YouTube
AI Exposes Major PHP Flaw: Global Websites Go Offline FOX 35 Orlando · YouTube
AI Found Critical PHP Flaws: Up to 70% of Websites at Risk, Patch Now #Shorts ACloudLabs · YouTube
AI Found a 14-Year-Old Hole in the Code Behind 70% of the Web Adem B10 · YouTube
PHP rocked by critical 9.2 exploit… and AI found it DEV-WRAP · YouTube
PHP Vulnerability: AI Found a Hole in 70% of Websites (and 3 More AI Stories) Simply AI · YouTube
Millions of Cyberattacks Target Critical PHP Software Flaws Worldwide B.C. Begley · YouTube
What Happens Now? #PHP2Shell, the Bug in 70% of the Internet Starving Piggy Productions · YouTube
PHP2Shell: AI uncovers critical RCE vulnerabilities affecting 70% of the Internet. #phpsecurity . Creative Rajat · YouTube
OpenAI and Google’s Wiz Helped Uncover Critical PHP Flaws AI Orbit · YouTube
PHP2Shell AI Uncovers Critical PHP Vulnerabilities Affecting Seventy Percent of the Web Eddy Says Hi #EddySaysHi · YouTube
“We Are Very VULNERABLE!” | Websites Taken Offline After AI Discovers Major Glitches The Julia Hartley-Brewer Show on Talk · YouTube
70% of the Web Had a 14-Year-Old Flaw Numbers, Visualized · YouTube
PHP2Shell: AI helped find 2 critical PHP flaws. Patch now Hit list AI · YouTube
⚠️ 用 PHP 嘅網站即刻更新!OpenAI / Google AI 聯手發現 9.2分超高危漏洞! Today in AI · YouTube
🚨 CRITICAL PHP SECURITY ALERT: 70% OF THE WEB COULD BE AT RISK! 🌐🔐 Gift-Tech Solutions · YouTube
AI Found a Critical Flaw in PHP😳 #shorts Your AI Sakha · YouTube
🚨 全球網頁大地震!AI 揪出 PHP 20年史詩級漏洞:CVSS 達 9.2 分! Today in AI · YouTube
PHP patches two critical remote code flaws that AI helped find GenAI Under the Hood · YouTube
Is Your WEBSITE OK? - OpenAI Helped Uncover a Major Web Security Flaw — What It Means for You Dr Jim Kennedy · YouTube
Why Governments Suddenly Shut Down Their Own Websites (PHP) The Cyber AI Files · YouTube
A Master Key to the Internet Was Just Found Omiro · YouTube
PHP2Shell: AI Helped Find a Flaw Behind 70% of Sites Cost of Intelligence · YouTube
IS YOUR WEBSITE OK??? PHP ALERT - OpenAI Helped Uncover a Major Security Flaw What It Means for You Dr Jim Kennedy · YouTube
PHP2Shell: Why Websites Went Offline on Purpose Proof & Pattern · YouTube





Two critical vulnerabilities in PHP, the programming language underlying approximately 70% of the world's websites, were disclosed on Wednesday, October 23, 2026, triggering emergency response measures across government and private sectors worldwide. The flaws, officially designated CVE-2026-109618 and CVE-2026-109619 and collectively termed "PHP2Shell," enable remote code execution and unauthenticated access to affected systems without requiring a password. Researchers at Calif and Wiz, working with OpenAI's Patch the Planet initiative, discovered methods to exploit vulnerabilities that had remained dormant for years — with one flaw originating from a warning issued in July 2004.
The crisis unfolded rapidly after public disclosure. Cloudflare CEO Matthew Prince reported that in the 10 hours before the patch was released, the company's network detected thousands of exploitation attempts; this number escalated to tens of thousands within 45 minutes of the announcement, hundreds of thousands within two hours, and millions by Thursday. Despite this activity, cybersecurity firms including Sophos, Palo Alto Networks, CrowdStrike, and others reported that as of late Thursday, October 22, they had not detected successful attacks in the wild. Sophos vice president Don Smith noted that "hackers were finding it difficult to capitalize on the vulnerabilities," partly because the flaws proved technically complex to exploit. Palo Alto Networks' Sherrod DeGrippo stated that over 200 proofs-of-concept exploit code examples were identified, but whether active exploits would materialize remained a question of time.
Governments and critical infrastructure operators worldwide responded by either patching systems urgently or taking websites offline. Major websites and services including Wikipedia (via Wikimedia Foundation), government portals in Finland, the Netherlands, Scotland, and the United Kingdom, banking services in Japan including those operated by Mitsubishi UFJ Financial Group and Nomura, transportation networks such as East Japan Railway and ANA, and numerous retailers and gaming platforms temporarily shuttered operations. Japan's SoftBank, the nation's third-largest mobile carrier, suspended its emergency alert system for mobile users — a significant concern given Japan's vulnerability to earthquakes and typhoons. The National Security Agency (NSA), UK National Cyber Security Centre (NCSC), FBI, and cybersecurity authorities in Europe and Australia issued coordinated advisories urging immediate patching or temporary closure.
The vulnerability discovery marked a watershed moment in AI-assisted security research. Researchers used advanced AI models to analyze PHP code and identify exploitation paths that would have required weeks or months of manual labor. OpenAI noted that while AI accelerated the discovery process, this capability also raised concerns that malicious actors could use similar tools. However, Calif researcher Ryan Austin clarified that the company used only publicly available AI models. The rapid publication of technical details forced the PHP Foundation to accelerate public disclosure on Wednesday morning, accompanied by patches for multiple affected PHP versions released since the mid-2000s.
As of late Thursday, global security monitoring revealed few confirmed successful exploits, though experts warned the threat remained acute. Vyas Sekar, computer engineering professor at Carnegie Mellon University, characterized the disclosure as "a wakeup call," noting that AI democratization had eliminated "security by exhaustion" — the prior reliance on vulnerabilities being too tedious to exploit without automation. Japan announced plans to establish a team to simulate cyberattacks using AI to identify remaining weaknesses. Cybersecurity researchers cautioned that thousands of systems worldwide remained unpatched and vulnerable, particularly small organizations and sites with limited security resources. The incident underscored the dependency of modern internet infrastructure on aging open-source technologies maintained by small teams.
Why This Matters
Two critical remote-code-execution vulnerabilities in PHP affect approximately 70% of websites globally. Within hours of October 23, 2026 disclosure, millions of exploitation attempts were detected across global networks; no confirmed successful breaches were reported as of late October 23. Immediate consequences: Wikipedia, government portals (Finland, Netherlands, Scotland, UK), Japanese banks (Mitsubishi UFJ, Nomura), railways (East Japan Railway, ANA), SoftBank's emergency alert system, and numerous retailers temporarily went offline for emergency patching. Organizations without patches remain at risk; security researchers warn that thousands of systems worldwide remain unpatched, especially small organizations and sites with limited security resources. The incident has raised concerns about the vulnerability of aging open-source infrastructure and the dual-use implications of AI-assisted vulnerability discovery.
Timeline & Sources
Jan 1, 2020
WireRelated PHP vulnerability discovered; partial fix issued
Jan 1, 2024
WireSimilar PHP vulnerability uncovered but not patched
Oct 23, 2026
WirePHP Foundation releases security patches
Oct 25, 2026
WireJapan announces plan to establish cyberattack simulation team using AI
Related Signals
Sources
- PHP是什麼?為何它引發全球網絡安全危機?bbc_zhongwenMediaOct 23, 2026
- PHP-ocalypse: How two critical web bugs almost broke the internettechcrunchMediaOct 24, 2026
- Will AI expose the rickety infrastructure of the internet?techcrunchMediaOct 24, 2026
- Hackers appear to be struggling to exploit PHP flaws after global cybersecurity crisisNBC NewsMediaOct 22, 2026
- AI识别PHP重大漏洞,多国政府和企业网站临时关停nytimes_cnMediaOct 23, 2026
- The biggest-ever internet security threat has been known for 22 years. AI made it a crisis in 2 daysCNNMediaOct 23, 2026
- Japan to simulate cyberattacks as PHP flaws roil global internetThe Express TribuneMediaOct 25, 2026