Case Study · Replit-based VoiceAgent
Replit-Built Apps Are Fast to Ship… and Even Faster to Exploit

AI-coded App Left 59 Open Vulnerabilities, and Open Access to Critical Interview Data

COMPANY DESCRIPTION

VC-backed AI VoiceAgent platform built entirely using Replit's AI coding assistant to help professionals prep for stakeholder interviews using voice-based simulations, autogenerated documents, and smart prompt creation.

Perfai Security's Role

Autonomous app testing across 71 endpoints. Exposed paths including live interview transcripts, project documents, Replit prompt generators, and patchable interview notes. All flagged and auto-fixed by Perfai Security's agentic models.

App Count
1
Endpoints
71
Vulnerabilities
59
Critical
25
High
6

What we found, in plain English

Full technical findings

Exploitable Surfaces & Risks

  • User-Uploaded Documents
    Sensitive data including PII
  • Interview Sessions
    Metadata, schedules, participant names
  • Internal Logs and Coaching data
    Personalized advice and sensitive user context
  • Direct Leakage of Prompt Engineering Logic
    Critical intellectual property and prompt chains
  • Transcripts of Voice Interviews
    Reputational, legal, and compliance implications
  • Privilege Escalation
    Altering and injecting data
  • Live or Stored Voice Conversations
    Violating user trust and voice privacy

Perfai Security Testing Categories

  • OWASP API Security Top-10 List
  • OWASP Web Top-10 List
  • Perfai Security Next-Gen Top-30 List
  • OWASP Business Logic Top-10 List
  • OWASP Mobile Top-10 List
  • OWASP AI Agent Top-10 List
Estimated Breach Cost With:
50 Customers | $455,000

See what Perfai Security finds in your app

Paste a URL. Get a free security report in minutes.

Start Free →