Google halting cyberattack with AI detecting zero-day vulnerabilities, highlighting global AI-driven threats from China and N

Editorial illustration for Google stops attack after AI finds zero‑day; China, North Korea also using AI

Google stops attack after AI finds zero‑day; China,...

Google stops attack after AI finds zero‑day; China, North Korea also using AI

By AI Daily Post Edited by Brian Petersen, Editor-in-Chief

May 12, 2026 • Updated: May 14, 2026 • 2 min read

Google’s Threat Intelligence Group says it thwarted a mass cyber‑attack that hinged on an AI‑discovered zero‑day. While the exploit itself remains undisclosed, the report marks the first time a threat actor has been linked to AI‑driven vulnerability hunting at scale. The attackers allegedly weaponized the flaw before Google intervened, halting the planned onslaught.

State‑backed groups from China and North Korea are also putting AI to work, scanning code for weaknesses in a systematic fashion. One of the tools highlighted is “wooyun‑legacy,” a Claude plugin that draws on more than 85,000 real‑world vulnerability cases harvested from the Chinese platform WooYun, ostensibly to help AI models parse code more efficiently. Meanwhile, Russia‑linked crews are slipping AI‑generated obfuscation into malware such as the Android strain PROMPTSPY, which taps the Gemini API to steer infected devices.

Criminal outfits like “TeamPCP” are targeting the AI supply chain, eyeing popular open‑source packages. In response, Google has rolled out its own AI‑based defenses, including tools dubbed Big Sleep and CodeMender. The full GTIG report is available for deeper analysis.

Google says it stopped a mass cyberattack after AI was used to discover a zero-day exploit A new report from Google's Threat Intelligence Group (GTIG) details how attackers are using AI at scale for cyberattacks.

— Google, Google says it stopped a mass cyberattack after AI was used to discover a zero-day exploit - THE DECODER

Why this matters

Google’s Threat Intelligence Group says its AI tools identified a zero‑day that a threat actor was about to weaponize, prompting the company to halt a planned mass cyberattack. How many similar operations have already slipped past detection? The report also flags state‑backed groups in China and North Korea that are training AI to hunt for flaws, suggesting a shift toward automated vulnerability discovery at scale.

Yet the extent of their success remains uncertain; we’ve no data on how many exploits have been deployed beyond the one Google intercepted. The GitHub project “wooyun‑legacy,” described as a Claude plugin housing over 85,000 real vulnerability cases from the Chinese WooYun platform, illustrates how publicly available datasets can be repurposed for malicious ends. For developers, this underscores the need to scrutinize third‑party AI tools that ingest security data.

Founders should consider whether their threat models account for AI‑augmented attackers. Researchers must weigh the benefits of open vulnerability repositories against the risk of automated exploitation, a balance that is still being defined.