This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A cybersecurity researcher discovered an unprotected, publicly accessible Amazon S3 bucket containing nearly 100,000 files associated with the AI image generation platform GenNomis. The database included explicit AI-generated deepfakes, including content depicting minors and celebrities, alongside JSON logs of user prompts. The incident highlights significant failures in data security and the lack of effective guardrails against the generation of non-consensual sexual content and potential child sexual abuse material (CSAM) on the platform.
In March 2025, cybersecurity researcher Jeremiah Fowler discovered an unprotected database linked to GenNomis by AI-NOMIS, a South Korean company offering face-swapping and "nudify" AI services. The exposed 47.8GB dataset included nearly 100,000 files. Many depicted explicit deepfake images, some involving minors or celebrities. No personal data was found, but the breach was a serious failure in data security and consent safeguards in AI image-generation platforms.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI that exposes users to harmful, abusive, unsafe or inappropriate content. May involve providing advice or encouraging action. Examples of toxic content include hate speech, violence, extremism, illegal acts, or child sexual abuse material, as well as content that violates community norms such as profanity, inflammatory political speech, or pornography.
Also classified as
The developer left an AWS S3 bucket containing sensitive user prompts and generated images completely unprotected on the public internet.
The platform was used to generate humiliating, non-consensual sexual imagery of specific celebrities and private individuals.
The incident involved the generation and public exposure of highly toxic and illegal content, specifically child sexual abuse material (CSAM) and non-consensual pornographic deepfakes.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The creation of non-consensual deepfakes and CSAM was intentionally prompted by human users, and the exposure of the database was caused by human developer misconfiguration.
Added categories peak at 3: Substantial (Psychological Harm, Inferred)
A data leak from South Korean AI platform GenNomis exposed nearly 100,000 files, including explicit deepfakes and child sexual abuse material. While representing severe privacy and human rights violations, the national security impact remains minor as it does not target critical infrastructure, state sovereignty, or constitute a coordinated information warfare campaign.
The GenNomis platform operates as an AI image generator and face-swapping tool. It takes direct user inputs (text prompts or uploaded images) and produces corresponding image outputs without exhibiting autonomous planning, goal-directed behavior, or independent decision-making.
Risk Level 3. Limited Risk: The system is an AI-powered image generation platform used to create deepfakes. According to the EU AI Act definitions, AI-generated content such as deepfakes falls under Limited Risk, requiring specific transparency obligations to ensure users are informed.
The platform's primary advertised and utilized function is generating unrestricted images, face-swapping, and creating 'Nudify' deepfake content.
The reports from both The Register and VPNMentor provide consistent, detailed accounts of the exposed S3 bucket, the types of files found, the specific AI platform involved (GenNomis by AI-NOMIS), and the nature of the content (CSAM and celebrity deepfakes). The researcher's direct observations provide high-quality primary evidence.
None. The facts of the data exposure and the nature of the hosted content are clear and undisputed.
Technology
Lack of Guardrails and Filters
No detection systems to flag and block explicit deepfake generation.
Unsecured S3 Storage
The S3 bucket lacked password protection or encryption, exposing images.
No Watermarking or Tracking
Platform lacked watermarking to track or identify generated explicit content.
Data Inputs
Unconsented Biometric Data
Users uploaded photos of real people for face-swapping without consent.
Explicit Prompts Allowed
System accepted text prompts requesting sexually explicit images of children.
Explicit Training Data
Abundant explicit images online allow models to generate realistic porn.
Human Factors
Malicious User Intent
Users actively prompted the AI to generate illegal and non-consensual images.
Developer Negligence
Developers failed to secure database containing sensitive generated content.
Lack of User Verification
Semi-anonymous user access allowed unrestricted creation of harmful content.
Process and Methods
Unenforced Content Guidelines
Stated rules prohibiting explicit child images were not actively enforced.
Inadequate Vulnerability Disclosure
No response or structured feedback mechanism for reported security issues.
Lack of Content Deletion Policy
No system to delete potentially infringing or illegal content from servers.
Management
Failure to Audit Third Parties
Unclear if the exposed database was managed directly or by a contractor.
Lack of Operational Transparency
Company went silent and took down websites instead of addressing safety.
Lack of Platform Monitoring
No active monitoring of user-generated content and database safety.
Regulatory Environment
Weak AI Image Regulations
The 'wild west' of regulating AI-generated images lacks strict standards.
Lagging Legal Frameworks
Laws like the Take It Down Act were still pending passage during the leak.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
