ReadingResearchRadarInvestment framework
Sign in / Sign up
← Back to reading
TopicDisconfirmAI EditorialUpdated 2026-11-21 20:39

Shopify AI Security

4 steps1 pieces4 Sources

Shopify has fundamentally restructured its mobile application security protocols to cope with an 11-fold increase in code throughput driven by agentic development. By November 2026, the company was shipping approximately 434,000 lines of code per week for its Shopify app, compared to 39,000 lines a year prior. To manage this volume without sacrificing safety, Shopify implemented a system centered on "attack surface maps" and specialized AI skills. These maps inventory every security-sensitive component—such as deep links, WebViews, and stored data—and are reviewed daily by frontier models that assess changes against OWASP standards and historical incident lessons.

The approach extends beyond passive scanning. In July 2026, Shopify detailed an agentic harness named Dispatch that not only discovers vulnerabilities but proves them with real tests and generates Shopify-tuned fixes. This system has scanned over 80 applications, including the massive Rails monolith, producing findings valued at over $400,000 in equivalent bug bounty payouts. Complementing this is the Cyber Chaos Monkey (CCM), introduced in November 2026, which uses AI agents to actively attack production infrastructure using live credentials to identify weaknesses before adversaries can exploit them. The CCM has executed tens of thousands of offensive engagements without causing outages.

Remediation is also automated through River, an AI agent integrated into Slack that drives vulnerability fixes from detection to merge. Since its operationalization, the backlog of open dependency issues fell by about 70% in the first 11 days, and security merges through their freshness-gated queue rose from 10% to 80%. For third-party developers, Shopify provides tools like shopify app security check to integrate these practices into CI/CD pipelines, enforcing strict deadlines for IP allowlists and expiring tokens starting in late 2026.

How it got here

  1. 2026-07-29

    Jul 29, 2026: Shopify published details on 'Dispatch', an agentic code review harness that discovers vulnerabilities, proves them with tests, and provides fixes. It has scanned over 80 apps, yielding findings valued at >$400k.

Sign in / Sign up
ReadingResearchRadarInvestment framework
shopify.engineering ↗
  • 2026-09-02

    Sep 2, 2026: Shopify introduced 'River', an AI agent in Slack that automates vulnerability remediation from patch to merge. Early results showed a 70% drop in open dependency issues and an increase in security merges to 80%.

    shopify.engineering ↗
  • 2026-11-12

    Nov 12, 2026: Shopify announced 'Cyber Chaos Monkey' (CCM), an AI-driven offensive security tool that attacks production infrastructure to find bugs. It has conducted tens of thousands of engagements without incidents.

    shopify.engineering ↗
  • 2026-11-18

    Nov 18, 2026: Michael Hughes described how Shopify uses 'attack surface maps' and AI skills to manage security reviews amidst an 11x increase in code output (~434k lines/week).

    shopify.engineering ↗
  • What is still unsettled

    • How does Shopify quantify the reduction in false positives specifically attributable to the 'test oracle' verification step in the Dispatch harness?
    • What are the specific criteria for when the Cyber Chaos Monkey escalates a finding to human engineers versus auto-remediating it?
    • How do the new security deadlines for third-party apps (IP allowlists, expiring tokens) impact the developer ecosystem's adoption rate?

    Sources

    • shopify.engineeringBuilding an agentic harness that outlasts the model↗
    • shopify.engineeringHow River takes security work from a fix to merge↗
    • shopify.engineeringCyber Chaos Monkey: Staying ahead of security threats with AI agents↗
    • shopify.engineeringBuilding secure mobile apps in the agentic era↗

    Coverage in this topic

    • When AI Pushes Code Output Up 11x, Shopify Uses Attack Surface Maps to Break Security Reviews into Small Questions2026-11-19News

      Michael Hughes of the Shopify engineering team describes their approach for mobile apps: maintaining an attack surface map grouped by surface, writing OWASP standards and lessons from past incidents into updatable skill files, reviewing changed entries in the map daily, and reading through change files for omissions. This keeps security reviews manageable despite a throughput of approximately 434,000 lines of code changes per week.