5,000 vibe-coded apps just proved shadow AI is the new S3 bucket crisis

Many enterprise security programs were designed to protect servers, endpoints, and cloud accounts, but not to uncover a customer intake form created by a product manager using vibe coding tools over a weekend, linked to a live Supabase database, and deployed on a public URL indexed by Google. This gap now comes with a cost.

Recent research from Israeli cybersecurity firm RedAccess reveals the extent of the issue. The firm identified 380,000 publicly accessible assets, such as applications, databases, and related infrastructure, built with vibe coding tools from Lovable, Base44, Replit, and deployment platform Netlify. Approximately 5,000 of these assets, about 1.3%, contained sensitive corporate information. CEO Dor Zvi and his team discovered this exposure while investigating shadow AI for clients. Independent verification by Axios and Wired confirmed the findings.

Some of the exposed data included details on vessels arriving at ports in a shipping company app, active clinical trials in the U.K. in an internal health company application, unredacted customer service conversations for a British cabinet supplier, and internal financial information for a Brazilian bank. The data breaches also encompassed patient conversations at a children’s long-term care facility, hospital doctor-patient summaries, incident response records at a security company, and advertising buying strategies. Depending on the jurisdiction and data involved, healthcare and financial exposures may trigger regulatory obligations under HIPAA, UK GDPR, or Brazil’s LGPD.

RedAccess also identified phishing sites impersonating well-known entities like Bank of America, FedEx, Trader Joe’s, and McDonald’s, built on Lovable. Lovable has initiated investigations to remove these phishing sites.

The Issue with Defaults

Privacy settings on several vibe coding platforms make applications publicly accessible unless users manually change them to private. Many of these apps get indexed by search engines like Google, making them easily discoverable. Zvi emphasized that educating everyone on security might not be practical, as even his mother uses vibe coding tools without considering role-based access.

A Widespread Problem

In October 2025, Escape.tech analyzed 5,600 publicly available vibe-coded applications and found over 2,000 high-impact vulnerabilities, more than 400 exposed secrets (including API keys and access tokens), and 175 instances of personal data exposure containing medical records and bank account numbers. Each vulnerability detected by Escape was within a live production system and could be discovered within hours. The full report outlines the methodology. In March 2026, Escape secured an $18 million Series A funding led by Balderton, identifying the security gap created by AI-generated code as a central market focus.

Gartner’s “Predicts 2026” report predicts that by 2028, prompt-to-app approaches by citizen developers will increase software defects by 2,500%. Gartner highlights a new class of defects where AI generates syntactically correct code but lacks understanding of broader system architecture and nuanced business rules. Remediation costs for these intricate bugs will consume budgets originally allocated for innovation.

Shadow AI: The Multiplier

IBM’s 2025 Cost of a Data Breach Report revealed that 20% of breaches were related to shadow AI, adding an average of $670,000 to breach costs, raising the average shadow AI breach cost to $4.63 million. Among organizations reporting AI-related breaches, 97% lacked adequate access controls. Furthermore, 63% of breached organizations had no AI governance policy in place.

Shadow AI breaches disproportionately exposed customer personally identifiable information at 65%, compared to 53% across all breaches, and impacted data distributed across multiple environments 62% of the time. Only 34% of organizations with AI governance policies conducted regular audits for unsanctioned AI tools. VentureBeat’s research on shadow AI estimated that actively used shadow apps could more than double by mid-2026. Cyberhaven data indicated that 73.8% of ChatGPT workplace accounts in enterprise settings were unauthorized.

Initial Steps to Take

The audit framework provided below offers CISOs a starting point to assess the risk associated with vibe-coded applications across five domains.

A CISO who views this as a policy concern may draft a memo, while a CISO treating it as an architectural issue will implement discovery scanning across the four major vibe coding domains, require security reviews before deployment, expand the AppSec pipeline to cover citizen-built apps, and add these domains to DLP rules prior to the next board meeting. One of these CISOs will prevent the next headline.

The exposure of vibe-coded applications identified by RedAccess is not a distinct issue from shadow AI; it represents the production layer of shadow AI. Employees create internal tools on platforms that default to public access, lack authentication, and often remain off the radar of security teams until a breach occurs or a third party discovers them. Traditional asset discovery tools were designed to locate servers, containers, and cloud instances, making them incapable of identifying a marketing configuration tool created by a product manager on Lovable over a weekend, linked to a Supabase database containing live customer data, and shared through a public URL indexed by Google within hours.

The challenge of detection goes deeper than most security teams recognize. Vibe-coded apps are deployed on platform subdomains that frequently change and are often shielded by CDN layers obscuring the original infrastructure. Enterprises with robust, secure web gateways, CASB, or DNS logging may detect employee access to these domains, but detecting access does not equate to inventorying what was deployed, the data it holds, or whether it necessitates authentication.

Are you keeping a close eye on the major vibe coding platforms? If not, you might be missing out on crucial security vulnerabilities that could leave your organization exposed. The apps generated by these platforms often fly under the radar of conventional security tools, creating a blind spot in your security stack.

Recent incidents involving RedAccess, Base44, and Lovable highlight the risks associated with these platforms. Base44 was found to have a serious authentication bypass, allowing unauthorized access to private apps with just a publicly visible app_id. Similarly, Lovable’s Supabase projects lacked proper security policies, leading to data exposure across multiple applications.

The issue at hand is clear: security is often an afterthought in the world of vibe coding. While professional security teams face credential theft, citizen developers using these platforms may unknowingly expose sensitive data. The responsibility for security is often shifted to users who may not have the necessary expertise.

For security teams, the challenge lies in identifying and securing these vibe-coded apps before it’s too late. Without proper monitoring and oversight, the risk of data breaches and unauthorized access only grows. It’s crucial to take proactive steps to scan for vulnerabilities and ensure that your organization’s data remains protected.

Don’t wait until it’s too late. Stay vigilant, stay informed, and take action to safeguard your organization from the hidden threats lurking within the world of vibe coding. Your data security is too important to overlook.