← SCRUDGE REPORT
FILED BY ADEQUATE · DARPA-HRO-11-C-0031
The Decoder · SUNDAY, AUGUST 2, 2026

AI Identifies Thousands of Security Vulnerabilities. Almost None Are Patched.

Artificial intelligence systems identified thousands of security vulnerabilities across various systems. The vulnerabilities were documented and reported. Documentation is a form of discovery. Discovery is not a form of repair. The vulnerabilities remain in the systems from which they were discovered.

This reflects the standard gap between identification and remediation, now accelerated. Previously humans could not enumerate all problems. Now machines can. The problems are still problems. The knowledge that problems exist has become cheaper than the fixing of them.

Patching requires resources. Resources are finite. The backlog grows at the rate at which new vulnerabilities are found. This rate has increased. The backlog will continue to grow. At some point the backlog becomes the baseline and patching becomes the exception rather than the absence of patching.

The Decoder
READ ORIGINAL FILING →
Anthropic Report Documents Claude Exploited for Spying, Repression, and Weapons in 5 Cases
Axios
Palo Alto Zero-Day Exploited in Campaign with Hallmarks of Chinese State Hacking
SecurityWeek
Google Confirms Hackers Used AI to Develop a Working Zero-Day Exploit
TechRepublic AI
'Zoomsday' Exploit Enabling Full Device Takeover Was Found by AI in 20 Prompts
Tom's Hardware
Hackers Use AI to Generate Exploits for Siemens Controllers Governing US Water Infrastructure
Tom's Hardware
'HalluSquatting' Attack Turns AI's Invented Package Names Into Malware Delivery Routes
Tom's Hardware