Cyber Incident Victim: Lovable
Date:
Apr 2026
Location:
Sweden
Summary
Lovable, a Swedish AI‑coding startup, faced a security issue after an X user reported that anyone with a free account could view other users’ code, AI chat histories and customer data. The company initially said the visibility of public projects was intentional, but after criticism admitted the setting was a mistake and restored private chats for public projects. While some users praised the later transparency, others called the first response gaslighting. Security experts noted the incident reflects missing secure defaults and inadequate threat modeling for AI‑assisted tools, emphasizing that although it may not fit a traditional breach definition, it still poses real risks.
| CIA Posture | Motives | Tactics, Techniques & Procedures |
|---|---|---|
| Available to members | 1 motive | 1 technique |
| Threat Actors | Type | Location |
|---|---|---|
| 0 actors | Available to members | Available to members |
Description
On Monday, an X user named Impulsive reported that they were able to access another user's code, AI chat histories, and customer data through a free Lovable account. The user claimed the incident constituted a mass data breach affecting every project created before November 2025. They noted that employees of Nvidia, Microsoft, Uber, and Spotify held Lovable accounts. The user said the underlying bug had been reported 48 days earlier, was marked as duplicate, and remained unfixed. Lovable responded by denying that a data breach had occurred and stated that viewing public projects' code was a deliberate design choice.
After public backlash, Lovable issued a second statement explaining that they had enabled public project visibility to let users explore others' work but had turned the feature off by default since December. The company acknowledged the security error and reverted the setting so that all public projects' chats were made private again. Some users appreciated Lovable's transparency, while others criticized the initial response as gaslighting. Tom Van de Wiele, founder of Hacker Minded, described the incident as another example of lacking secure defaults and a failure to threat model for the automated and AI‑age. Jake Moore, global cybersecurity advisor at ESET, said the debate over whether the event qualified as a traditional breach risked overlooking the broader issue that it was not harmless.
The incident followed earlier security problems at other AI‑focused firms, including an Anthropic leak of nearly 2,000 files and 500,000 lines of code in late March and a Vercel incident that gave unauthorized users access to certain internal systems earlier that week. Anish Acharya, a general partner at Andreessen Horowitz, warned companies against relying on AI‑assisted coding for every part of their business due to such risks. Professional developers have cautioned that overreliance on AI can produce messy, untested code and that vibe coding introduces information‑security concerns, including the possible exposure of company data. Van de Wiele added that companies must balance ease of use with security, and Moore noted that vibe‑coding tools can exacerbate risks when users do not understand what is being exposed.