GenNomis AI Database Reportedly Exposes Nearly 100,000 Deepfake and Nudify Images in Public Breach

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.

Risk classification

• Primary risk domain: 1 Discrimination & Toxicity
• Primary risk subdomain: 1.2 Exposure to toxic content

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.

Additional risk subdomains

• 4.3 Fraud, scams, and targeted manipulation: The platform was used to generate humiliating, non-consensual sexual imagery of specific celebrities and private individuals.
• 2.2 AI system security vulnerabilities and attacks: The developer left an AWS S3 bucket containing sensitive user prompts and generated images completely unprotected on the public internet.

Causal factors

• Entity: Human
• Intent: Intentional
• Timing: Post-deployment

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.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

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.

AI system and alleged parties

• AI system: GenNomis (AI-NOMIS)
• AI purpose: Deepfake Image Generation; Image Retouching
• Behaviour type: Tool
• Alleged developer: GenNomis by AI-NOMIS, AI-NOMIS
• Alleged deployer: GenNomis by AI-NOMIS, GenNomis
• Alleged harmed parties: Public figures and celebrities depicted in explicit AI images, minors, Individuals whose likenesses were used without consent, General public

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Substantial. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Minor, indirect Negligible
• Differential treatment: direct Minor, indirect Negligible
• Civil rights: direct Minor, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Substantial, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Minor, indirect Negligible

Malicious content

Reported: Yes, the report explicitly describes the creation and exposure of nearly 100,000 explicit images, including CSAM and non-consensual deepfakes of celebrities and other individuals.

Directly caused: The GenNomis platform was used to generate 93,485 explicit images, which were stored in an unprotected S3 bucket and exposed to the public internet.

Inferred additional harm: It is likely that similar unrestricted AI generation platforms continue to generate and expose massive volumes of toxic, non-consensual, and illegal explicit content across the internet.

Differential treatment

Reported: Yes, the report notes that almost all of the celebrities and individuals targeted for non-consensual deepfakes were women, reflecting a gendered pattern of abuse.

Directly caused: The platform facilitated the generation of sexually explicit deepfakes targeting women and female celebrities (e.g., Kardashians, Beyonce, Michelle Obama, Kristen Stewart).

Inferred additional harm: This reinforces systemic gender-based harassment and the disproportionate targeting of women with non-consensual sexual imagery.

Civil rights

Reported: Yes, the report describes violations of personal privacy, dignity, and the rights of children regarding sexual exploitation.

Directly caused: The generation of CSAM and non-consensual pornography violates the fundamental rights to dignity, privacy, and protection from sexual exploitation of the depicted minors and adults.

Inferred additional harm: Widespread availability of such tools threatens the basic civil rights of individuals to control their own likeness and be free from sexual harassment.

Privacy

Reported: Yes, the report describes the exposure of user prompts, links to generated images, and uploaded photos of real people used for face-swapping without consent.

Directly caused: An unprotected AWS S3 bucket exposed 93,485 images and JSON files logging user prompts and links, violating the privacy of both the users and the non-consenting individuals whose likenesses were used.

Inferred additional harm: It is highly likely that the biometric data and personal likenesses of thousands of individuals were compromised and potentially accessed by unauthorized third parties during the period the database was exposed.

Child sexual exploitation and abuse

Reported: Yes, the report explicitly describes the discovery of AI-generated explicit images of children and celebrities portrayed as children within the exposed database.

Directly caused: The GenNomis platform was used to generate explicit images of children and celebrities portrayed as children, which were then stored in the publicly exposed database.

Inferred additional harm: The existence of these files suggests that users actively use such unrestricted platforms to generate CSAM, posing a severe ongoing risk to child safety.

People affected

• Occurrences reported: 1
• People reportedly harmed: 10
• People reportedly exposed: 100

Potential causes

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.

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.

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.

Information quality

• Classification confidence: High
• Reason for confidence: 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.
• Ambiguities identified: It is not fully clear how many unique users generated the content or how many distinct non-consenting individuals had their faces swapped.
• Alternative interpretations: None. The facts of the data exposure and the nature of the hosted content are clear and undisputed.

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.

• Overall national security impact: Minor
• Scope: Multiple nations
• Primary target: No clear primary
• Other affected: United States, South Korea
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. Represents an ongoing strategic concern regarding the abuse of generative AI rather than an active, imminent national security crisis.
• Autonomy: Human-controlled. The AI system acted as a tool directly driven by human prompts and uploads without autonomous decision-making capabilities.
• Novelty: Established threat. The use of generative AI for creating deepfakes and the exposure of unsecured S3 buckets are well-established threat vectors.

Impact by dimension

• Physical security: Negligible. No physical systems, critical infrastructure, or kinetic targeting capabilities were affected or compromised in this incident.
• Information security: Negligible. While the database contained deepfakes of public figures, there is no evidence of state-sponsored information warfare or intelligence compromise.
• Sovereignty: Negligible. The incident does not involve threats to state authority, electoral systems, or core government operations.
• Economic security: Negligible. No strategic technology theft, financial system attacks, or critical economic infrastructure disruptions were reported.
• Societal stability: Minor. The exposure of CSAM and non-consensual deepfakes represents a significant human rights violation, though manageable under standard law enforcement frameworks.