■ TOP STORY

AI TURNS SECURITY PATCHES INTO EXPLOITS IN HOURS

Anthropic's Mythos Preview model can convert security patches for Firefox and Windows kernel into working exploits within hours for a few thousand dollars, with no specialized knowledge required. Eight complete attack chains were completed before Microsoft's auto-updates deployed.

► WHY IT MATTERS: The patch-to-exploit window has collapsed from weeks to hours, making traditional patch-and-update defenses obsolete and forcing enterprises to rethink vulnerability response timelines.

CISA MANDATES 3-DAY PATCH WINDOWS FOR FEDERAL AGENCIES

The Cybersecurity and Infrastructure Security Agency has directed US federal agencies to patch critical security vulnerabilities within as little as 3 days, citing AI-accelerated exploit development as the driving threat.

► Federal security mandates are explicitly tightening response times due to AI capabilities, signaling a structural shift in how government treats vulnerability management across all critical infrastructure.

COUPANG FINED $409M FOR 33.7M-ACCOUNT BREACH

South Korea's largest e-commerce platform, Coupang Inc., was hit with a record 624 billion won ($409M) fine for a 2025 data leak exposing approximately 33.7 million accounts, which escalated into a diplomatic conflict between South Korea and the US.

► Record-sized penalties and cross-border diplomatic fallout signal that regulators are weaponizing data breach fines as enforcement tools, not just compliance costs.

SHINYHUNTERS BREACHES ORACLE PEOPLESOFT SERVERS

The ShinyHunters extortion gang is actively targeting Oracle PeopleSoft servers in ongoing data theft attacks, claiming to have stolen data from over 100 organizations.

► Enterprise resource planning platforms remain high-value targets for extortion gangs, and PeopleSoft's continued exposure indicates that older legacy ERP infrastructure is particularly vulnerable.

SPACEEX ALUMNI STARTUP TARGETS AI POWER CRISIS

Ambrosia Energy, founded by SpaceX alumni, is building solar and battery power plants in under 12 months at costs undercutting natural gas, with plans to deploy gigawatts by 2030 to address AI infrastructure energy demands.

► AI's power consumption crisis is attracting top-tier engineering talent away from aerospace into energy infrastructure, suggesting energy—not chips—may become the binding constraint for AI scaling.