everyone JUST got HACKED...

TL;DR

AI models like Claude Opus and GPT-5.5 are accelerating exploit discovery at unprecedented rates, with experts warning a 'bug apocalypse' is imminent.

Key Points

• 1.Claude Opus helped break Apple's decade-long memory integrity enforcement in five days. Security researchers at Khalif.io used Opus (via Anthropic's Project Glasswing) to find a kernel privilege escalation exploit on macOS 15.4.1 on M5 hardware, delivering a 55-page physical report to Apple's Cupertino HQ.
• 2.Google stopped the first confirmed AI-assisted mass exploitation event. On May 11th, Google Threat Intelligence discovered attackers used an LLM to build a working zero-day exploit targeting two-factor authentication in a popular open-source web admin tool — identified as AI-generated because the malware script included a hallucinated CVSS severity score.
• 3.OpenClaw was flagged by Google as a tool used by attackers. Google's blog post named OpenClaw and OneClaw as agentic tools used by threat actors to develop and refine AI-generated exploits in controlled environments before deployment.
• 4.Microsoft's M-dash system — 100+ models in orchestration — now outperforms both Claude Opus and GPT-5.5 on cybersecurity benchmarks. This supports the argument that multi-model orchestration beats single large models, a position Microsoft has held since Opus launched.
• 5.Palo Alto Networks found 75 vulnerabilities in one month — seven times their normal rate — after gaining access to Claude Opus and GPT-5.5 Cyber. They estimate a 3–5 month lead time before what they're calling the 'Vuln Apocalypse.'
• 6.Dario Amodei warned of a 6–12 month window to patch thousands of vulnerabilities before Chinese AI catches up. He cited risks of massive increases in ransomware targeting schools, hospitals, and banks; JP Morgan, Goldman Sachs, Citi, and others all have Glasswing access and are patching rapidly.
• 7.A Barracuda 'Mythos Hype Index' sits at 94/100, arguing the predicted CVE surge hasn't materialized. The video counters that zero-day exploits are being quietly patched without public CVE disclosure to avoid giving bad actors a roadmap, meaning absence from the CVE database doesn't mean bugs aren't being found.