UpTrajectory Review

AI-Generated Code: A Growing Security Risk for Businesses

robot writing code on glass screen
Photo: cottonbro studio / Pexels

UpTrajectory Review

Checkmarx's new survey of 2,350 security leaders across 14 countries lands at an awkward moment: enterprises now ship nearly half their production code as AI-generated, yet the same report shows that heavy AI adopters (those letting AI write 81% to 100% of their code) release vulnerable code 3.4 times more often than conservative users. The report's framing is blunt — enterprises relying on traditional security tools and methods 'cannot survive this reality' as AI models like Anthropic's Mythos collapse the vulnerability-to-exploit window from months to minutes. For readers who run lean dev teams or lean on contractors to ship AI-assisted code, this is not an abstract enterprise problem. It is your release pipeline.

Why should a small-business operator care? Because the attack surface has shifted left and widened at once. Seventy percent of developers surveyed said AI code generation created vulnerabilities in 2025, and 93% of enterprises reported at least one breach tied to in-house apps. If you are a founder, ops lead, or the person who signs off on a SaaS build, you likely lack a dedicated AppSec team. You rely on the vendor's assurances, a pen test you commissioned once, or the hope that your stack is too small to target. The Checkmarx data says that hope is now a liability: the more AI code you ship, the more flaws you ship, and the faster those flaws become working exploits.

What is genuinely new here is not the existence of AI-generated vulnerabilities — that story is two years old — but the normalization of risk. Three-quarters of respondents, per the excerpt, treat AI code flaws as routine. That is a cultural failure, not a tooling gap. We are skeptical of one number in the report's framing: the 3.4x figure compares the heaviest AI adopters to the lightest, which may conflate correlation with causation. Shops that aggressively adopt AI may also cut review corners, skip threat modeling, or chase speed over safety. Still, the direction is clear and matches what we see in the field: AI code review is not keeping pace with AI code generation.

Second-order effects matter for procurement and hiring. If nearly half of production code is AI-written and most codebases lean heavily on open-source components, then your vendor's 'secure by design' claim is weaker than it was 18 months ago. Insurance carriers and enterprise buyers are starting to ask for software bills of materials (SBOMs) and evidence of AI code review in vendor assessments. Downstream, that means higher compliance costs, longer security questionnaires, and potential liability if you ship a product that leaks customer data because an AI-generated dependency had a known flaw. Developers also face workflow strain: they are expected to review AI output they did not write, often without training in secure code patterns.

What to do next: treat AI-generated code as untrusted input, not a productivity miracle. Require static analysis and dependency scanning before merge, not after deployment. Ask your vendors and contractors whether they use AI coding tools, what review gates they apply, and whether they can produce an SBOM on request. If you are evaluating a new build, budget for a security review that assumes AI involvement from day one. Watch for regulatory movement: as breach costs from AI-built apps climb, expect procurement standards and cyber insurance underwriting to tighten. The window between flaw and exploit is now minutes. Your review cycle needs to be faster.

“Mythos-class models collapse the window between a vulnerability existing and a working exploit being available from months to minutes” — CIO Magazine

Takeaway: Treat AI-generated code as untrusted input: require automated security scanning and dependency review before any merge or vendor handoff.

Excerpt from the original — CIO Magazine

AI-generated code is riddled with security flaws, yet enterprises are shipping more of it than ever before. Why? Perhaps they’re over-confident, lack true visibility into security risks, or are simply choosing to ignore the problem and hope it goes away.

It’s a dangerous game to play at the dawn of the agentic AI era, as underscored in a new report from app security company Checkmarx.

The survey of thousands of security leaders exposes an underlying naivete about AI-built code and its vulnerabilities, even as tools like Anthropic’s Mythos are uncovering security flaws orders of magnitude faster than any human security team could ever hope to.

“Mythos-class models collapse the window between a vulnerability existing and a working exploit being available from months to minutes,” the report notes. Enterprises relying on traditional security tools and methods, it says, “cannot …