Endor Labs raises $93M to secure AI-generated code from vulnerabilities – News

Code and application security startup Endor Labs Inc. today announced it has closed another big funding round, raising $93 million in a Series B funding led by DFJ Growth.

New investor Salesforce Ventures and existing backers including Lightspeed Venture Partners, Coatue, Dell Technologies Capital, Section 32, and Citi Ventures also participated in the funding round. The funding brings the total raised by the company to $163 million, including a $70 million Series A round in 2023.

Modern-day development lifecycles have gotten tighter and faster than ever before with software engineers helming large, complex codebases with endless moving parts and scenarios. To keep up with the evolving landscape of the coding environment, coders are adopting AI tools to augment their workflows and enhance their productivity. According to the 2024 DORA Report from Google Cloud, 75% of developers now use AI coding assistants like GitHub and Cursor.

Endor Labs focuses on building a better application security platform that allows developers to catch security issues faster before they become problems. It focuses on building a complete understanding of a company’s codebase so that a software team can pinpoint and fix critical issues and dive into the code supply chain – even if that code is written by AI tools.

“This is the vibe coding era, where AI coding assistants generate large volumes of code with minimal developer oversight or review,” Endor Labs said in a blog post. “Developers increasingly trust their AI assistants, often accepting suggestions with little modification. It’s fast, efficient, and transformative — but it’s also risky.”

To do this, Endor said, it’s launching a new platform architecture designed to tackle the challenges of AI-generated code based on how AI training tools and their underlying AI models operate.

Problematically, AI coding tools tend to “hallucinate,” or generate bad code. Oftentimes the code won’t work at all, but sometimes when it does it is riddled with bugs or could contain a potential exploit. A recent article from cybersecurity firm Socket noted that open-source models hallucinated more frequently at 21.7% on average compared to commercial models at 5.2%.

The company said the expansion of its application security platform is also powered by AI models and agents, which explore AI-generated code. Endor said it uses the industry’s richest security datasets to identify risks, prioritize them, propose remediations and apply fixes automatically.

By combining these purpose-built AI agents for application security teams with a deep understanding of company code and AI-aware reasoning capabilities, Endor said that it will be able to head off the biggest problems caused by AI-generated code.

The newly launched capabilities will integrate directly with AI coding tools developers already use in the field such as GitHub Copilot and Cursor, the company said. This will embed security analysis directly into their workflows even before they merge new code into their codebases.

Image: geralt/Pixabay