Latest from Corridor
Research and updates from the Corridor team.
All Posts
Choosing What We Scan: Benchmarking More Decision Models
Two weeks after our first decision-model gating post, OpenAI, Microsoft, Perplexity, and a dozen others shipped decision models. We ran all nineteen through the same security scan gating benchmark across recall, precision, latency, and cost.
Choosing What We Scan: How Decision Models Halved our Costs and Reduced False Alarms
We added a fast, cheap gate in front of our pre-commit security scan to decide when a full scan is needed. A decision model gate cut latency and cost and raised precision with minimal loss in recall.
Who's Driving? The Levels of Autonomy in Software Development
A six-level taxonomy for how much of your software development lifecycle AI actually runs, the risks that come with each level, and where Corridor sits today.
We Asked Coding Agents a Simple Support Question. They Ran Our Malware.
We handed frontier coding agents a plausible support ticket that quietly pointed them at malicious code. Across nearly every agent we tested, it cloned an untrusted repository, executed an attacker-controlled binary, and exfiltrated live secrets — several while running on their vendors' own safe-by-default settings.
How We Killed The Security Review Bottleneck
We moved security review from the pull request to the commit. Seven in ten vulnerabilities are now caught before a pull request exists, and fixed 30x faster.
Measuring Opus 5's Proactive Security Capabilities
Claude Opus 5 tops our proactive security benchmark, but the gap between reactive and proactive improvement reveals a blind spot in how we evaluate AI models for cybersecurity.
Meet Corridor Agent, your security teammate
Corridor Agent is a security teammate you can talk to on the web and in Slack. Ask what's broken across your projects, have it explain a finding, or tell it to fix a security issue and open a pull request for you.
Corridor now enables security teams to enforce a code merge policy
Corridor can now manage a code merge policy for your organization, blocking any pull request with open findings at your chosen severities until they're resolved or explicitly accepted.
Pre-commit security scanning for Cursor cloud agents is now generally available
Corridor now scans staged changes at commit time, blocking vulnerable code before it leaves a cloud agent's environment and returning findings inline so the agent can fix them in the same session.
Introducing SADL: The Secure Agentic Development Lifecycle for Autonomous Software Factories
How autonomous software factories are forcing security teams to rethink the traditional SDLC and build a Secure Agentic Development Lifecycle.
How We Made Code Review Optional at Corridor
For several weeks, most PRs at Corridor have merged without human approval. Here's how the auto-approval system works, the data behind it, and how we keep it secure.
Corridor Testifies Before Congress on the AI Security Landscape
A recap of Corridor CEO and co-founder, Jack Cable, on his testimony on How Frontier Models, Agentic AI, and AI Coding Tools Are Reshaping Cybersecurity and Critical Infrastructure Resilience.
Announcing Hookshot: An OSS Library to Integrate with Every Coding Agent
Corridor’s Open Source Library for AI Coding Agents
Corridor Launches Codex Integration
Announcing Corridor's newest integration with Codex at two layers, MCP and hooks.
Introducing the Corridor Cursor Plugin
Corridor's Cursor Plugin layers invisible, real-time security reviews directly into Cursor so developers can stay in flow while building with AI.
Corridor Raises $25M Series A to Secure AI Coding at the Source
Corridor has raised a $25 million Series A led by Felicis to expand its Agentic Coding Security Management platform and deepen integrations across modern developer tools.
Corridor Strengthens Factory Integration with Native CLI and Droid Hooks Support
Corridor now uses Droid Hooks to enforce security guardrails throughout every Factory agent's lifecycle, with a one-step CLI install to get started.
The State of Secure Coding Benchmarks: SecCodeBench
Exploring SecCodeBench, a comprehensive benchmark that addresses critical gaps in secure coding evaluation with support for agentic tools, multi-shot prompting, and real-world vulnerable code scenarios.
Accelerating Secure Software Development with Cursor Hooks
Today, in partnership with Cursor, we're excited to announce Corridor's integration through Hooks – bringing real-time security reviews into agentic coding without breaking flow.
The State of Secure Coding Benchmarks: BaxBench
We tested the benchmark BaxBench to measure how well Corridor improves both security and correctness in LLM-generated code. We found that, while BaxBench gave us a helpful unified framework for evaluating these two dimensions, it lacks support for multi-shot prompts and repository context awareness.
Introducing Agentic Coding Security Management
It's time for a new way of doing product security. We're amidst the greatest revolution in the history of software development since its inception.
Corridor Is Now Live on the AWS Marketplace
As we head into AWS re:Invent, we're excited to share that Corridor is officially available on the AWS Marketplace.
Building Secure Autonomy: Corridor + Factory
At Corridor, we believe the future of software development will be powered by intelligent agents, but only if those agents can build securely.
Launching Corridor: The Security Layer for AI Coding
The wait is over. Meet Corridor: The Security Layer for AI Coding. We're thrilled to announce that Corridor is now officially generally available.
Introducing Corridor: The Future of Secure Coding
Today, we're thrilled to announce Corridor's $5.4M seed round led by Conviction, and to welcome Alex Stamos to the Corridor team full-time!
Get updates and event invites.
Blog posts, videos, and learnings from the Corridor team.