AI-Accelerated Discovery of Decades-Old PHP Flaws Threatens Global Web Infrastructure

Two critical bugs in the PHP language, dormant for about two decades, have been thrust into the spotlight after AI‑driven analysis turned an obscure code path into an exploitable weapon. The flaws affect every PHP release since the early 2000s, covering an estimated 70 percent of the web, including platforms such as WordPress, Wikipedia, and numerous government services.

Researchers at Wiz reported that the vulnerability chain was previously at the edge of human discovery. Large language models and specialized reasoning engines reduced the effort from months of manual auditing to hours of automated reasoning, demonstrating how modern AI can accelerate both offense and defense in software security.

The fallout has been immediate: high‑traffic sites like DeviantArt and 4chan went offline, Dutch and Finnish agencies pulled public portals, and even Pokémon‑related services were suspended. Legacy applications that cannot be patched — such as decade‑old Japanese mobile games — face permanent retirement, erasing pieces of internet history.

Why it matters for GPU / AI infrastructure

The same GPU clusters that train foundation models are now being repurposed for large‑scale vulnerability hunting. Cloud providers must ensure their compute fleets can run intensive static‑analysis workloads securely, while customers need rapid patch deployment pipelines that leverage the same high‑performance hardware. AiGpu’s GPU‑as‑a‑service platform is positioned to supply the burst capacity required for both AI‑assisted research and emergency remediation.

Organizations should inventory all PHP‑based assets, prioritize supported versions, and allocate GPU‑accelerated scanning resources to detect exploitation attempts in real time. Proactive investment in AI‑enabled security tooling, backed by scalable GPU infrastructure, will reduce the window between disclosure and mitigation.