Checkmarx

Every board should ask how prepared the business is for a world where AI has lowered the barrier to sophisticated cyberattacks. Anthropic's new threat intelligence report (link in comment) makes that concrete: eight months of documented misuse, from state espionage to financially motivated fraud. The techniques nation-states once needed a team of specialists to run are now available to any criminal with an AI agent and a stolen API key. That's not a nation-state problem. It's a problem for every company with code and data worth stealing. As these models grow more capable, risk grows with them. That raises the bar for application security specifically. The code running organizations is where these threats land. To address it, AI companies and defenders must work together. That's why we continue to invest in security research through our Checkmarx Zero team, and why we recently joined Anthropic's Project Glasswing and launched Checkmarx Fusion. We're using Claude Mythos 5 to strengthen our own vulnerability detection. We're finding issues before anyone else does. And we're sharing what we learn with the industry. https://lnkd.in/gssVHRu7

Meeting slots at Les Assises are filling up. If you’ll be in Monaco 7–10 October, book 30 minutes with the Checkmarx team while there’s still availability. See how Developer Assist, Triage Assist, and Remediation Assist work together to help teams fix what matters at the speed of AI-driven development.

Heading to Les Assises in Monaco? Nearly half of AI-generated code contains vulnerabilities - and it's shipping faster than traditional AppSec can review it. That's the gap Checkmarx is built to close: hybrid scanning that keeps pace with AI-generated code, and AI agents that find, triage, and fix what it catches without slowing developers down. Book 30 minutes with our team at Les Assises to see it in action.

AppSec findings keep piling up. Fixes shipped keep falling. Your team isn't struggling because they can't find risk. They're struggling because they can't act on the findings fast enough.

Your team ran the frontier model to scan code. It found a lot, but the results changed from scan to scan. When developers can’t trust the findings, they stop acting on them.

Every language your scanner doesn't cover opens a door to risk. Every finding stuck in triage gives attackers more time to walk through it.