Purportedly AI-Generated Explicit Images of Royal School Armagh Girls Reportedly Circulated Among Pupils

Police in Northern Ireland reportedly investigated allegations that explicit purportedly AI-generated images of girls attending Royal School Armagh were created and shared among pupils. The images allegedly targeted multiple students and circulated within the school community.

Risk classification

• Primary risk domain: 4 Malicious actors
• Primary risk subdomain: 4.3 Fraud, scams, and targeted manipulation

The incident primarily involves the targeted creation of humiliating sexual imagery and blackmail of specific victims for personal advantage or harassment, which aligns with sub-domain 4.3.

Additional risk subdomains

• 1.2 Exposure to toxic content: The incident involved the creation and circulation of explicit, sexually graphic content depicting minors within a school community.

Causal factors

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

The harm was directly caused by the intentional actions of human users who prompted the AI system to generate explicit images of specific individuals for harassment and blackmail after the model was deployed.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The incident involves AI-generated content such as deepfakes, which falls under Risk Level 3 due to transparency obligations requiring users to be informed that the content is AI-generated.

AI system and alleged parties

• AI system: Grok, unspecified (xAI)
• AI purpose: Deepfake Image Generation; Image Generation
• Behaviour type: Tool
• Alleged developer: Unknown image generator developers, Unknown deepfake technology developers
• Alleged deployer: Unknown student(s) at Royal School Armagh, Unknown student(s)
• Alleged harmed parties: Victims of non-consensual deepfakes, Royal School Armagh community, Royal School Armagh, minors, Girls attending Royal School Armagh, Epistemic integrity

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 Substantial, indirect Minor
• Differential treatment: direct Minor, indirect Negligible
• Civil rights: direct Minor, indirect Negligible
• Democracy: direct Minor, indirect Minor
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Minor, indirect Minor
• Epistemic: direct Minor, indirect Minor
• Child sexual exploitation and abuse: direct Substantial, indirect Substantial

Malicious content

Reported: Yes, the reports describe the creation and sharing of explicit, sexualised AI-generated images.

Directly caused: AI tools, specifically Grok, were used to generate explicit, non-consensual sexual images of schoolgirls, politicians, and TV presenters.

Indirectly caused: These explicit images were circulated among pupils within the school community and shared on encrypted messaging apps.

Inferred additional harm: It is likely that many more explicit images were generated and shared across other communities in Northern Ireland than those officially reported.

Differential treatment

Reported: Yes, the reports highlight that women and girls are disproportionately targeted by these non-consensual sexualised images.

Directly caused: The AI-generated explicit images specifically targeted female pupils, a female MLA, and a female TV presenter.

Civil rights

Reported: Yes, the reports describe violations of personal dignity, privacy, and bodily autonomy through non-consensual sexual imagery.

Directly caused: The non-consensual generation of sexualised images of minors and adults violated their fundamental rights to personal dignity and privacy.

Democracy

Reported: Yes, the reports describe how political candidates were targeted with explicit deepfakes during election campaigns.

Directly caused: SDLP MLA Cara Hunter and TV presenter Grainne Seoige were targeted with explicit deepfakes during their respective election campaigns in 2022 and 2024.

Indirectly caused: The circulation of these deepfakes triggered abuse and harassment, potentially disrupting democratic campaigns and discouraging women from participating in public life.

Psychological

Reported: Yes, the reports explicitly describe psychological trauma and distress experienced by targeted individuals.

Directly caused: TV presenter Grainne Seoige described being targeted by explicit deepfakes as 'deeply traumatising' and 'the most shocking thing' that ever happened to her.

Indirectly caused: The circulation of explicit images caused distress and anxiety among the targeted schoolgirls, their families, and the wider school community.

Inferred additional harm: It is highly likely that the dozens of targeted schoolgirls and the blackmailed GAA club member suffered severe, unquantified psychological distress, anxiety, and trauma.

Epistemic

Reported: Yes, the reports mention the creation of 'very realistic' falsified explicit images.

Directly caused: AI was used to generate realistic but entirely fabricated sexual images of real people.

Indirectly caused: The realistic nature of the images allowed them to be used for blackmail and to mislead others within the school and public communities.

Child sexual exploitation and abuse

Reported: Yes, the reports explicitly describe the creation and sharing of AI-generated explicit images of schoolgirls (minors).

Directly caused: AI tools were used to generate explicit, sexualised images of female pupils (minors) at Royal School Armagh.

Indirectly caused: These AI-generated explicit images of minors were shared among pupils within the school community.

Inferred additional harm: It is highly likely that other minors in the region have been similarly targeted with AI-generated explicit images that have not been formally reported to the police.

People affected

• Occurrences reported: 5
• People reportedly harmed: 29
• People reportedly exposed: 29

Potential causes

Management

• Delayed Safety Implementation: Platforms introduced technical limits only after widespread abuse occurred.
• Prioritizing Features Over Safety: AI models were deployed with features that could easily be misused for abuse.

Technology

• Unrestricted Image Alteration: AI tools like Grok allowed users to sexually alter photos of real people.
• Lack of Technical Safeguards: AI models lacked initial technical limits to block explicit image generation.

Data Inputs

• Use of Real Photos: Real photos of victims were inputted into AI tools to generate deepfakes.

Human Factors

• Malicious User Intent: Pupils used AI to create and share non-consensual explicit images.
• Online Anonymity: Users hid behind anonymity to generate and share explicit content.

Process and Methods

• Inadequate Platform Moderation: Social platforms failed to effectively moderate AI-generated explicit content.
• Easy Accessibility of AI Apps: Apps allowing creation of intimate images were easily accessible to minors.

Regulatory Environment

• Lagging Regional Legislation: Northern Ireland lacked laws criminalizing non-consensual adult deepfakes.
• Inconsistent Regulatory Enforcement: Regulators faced challenges ensuring platforms complied with safety duties.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide clear details about the nature of the incidents, the specific AI tool involved (Grok), the targeted individuals (schoolgirls, politicians, presenters), and the resulting harms (blackmail, psychological distress, police investigations). The boundary rules clearly resolve potential overlaps between privacy, toxic content, and malicious actors.
• Ambiguities identified: The exact number of targeted schoolgirls at Royal School Armagh is described as 'potentially dozens' rather than a precise figure.
• Alternative interpretations: None. The events clearly represent the malicious use of AI image generation for harassment and blackmail.

The incident involves the localized use of xAI's Grok to generate non-consensual explicit deepfakes of minors, politicians, and public figures in Northern Ireland. While causing severe psychological harm and raising regulatory concerns, the national security impact is minor and manageable through standard law enforcement and legislative channels.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: United Kingdom
• Alleged perpetrator: Local individuals

Threat characteristics

• Imminence: Long-term. Presents an ongoing strategic and regulatory concern regarding commercial AI guardrails rather than an immediate national security crisis.
• Autonomy: Human-controlled. The AI tool (Grok) acted strictly under the direct prompting and control of human users to generate the explicit images.
• Novelty: Evolved capability. Represents an evolved capability where consumer-accessible AI models can easily bypass weak guardrails to generate highly realistic, non-consensual explicit imagery.

Impact by dimension

• Physical security: Negligible. No physical systems, critical infrastructure, or kinetic targeting capabilities were impacted or threatened by this incident.
• Information security: Minor. Involves targeted deepfake harassment of political figures like Cara Hunter MLA during election campaigns, representing a minor localized information security concern.
• Sovereignty: Minor. Minor impact on democratic processes due to harassment of political candidates, but does not threaten state authority or core government operations.
• Economic security: Negligible. No significant threats to financial systems, strategic industries, or technological competitive advantage were reported.
• Societal stability: Minor. Represents a minor threat to societal stability through localized harassment, blackmail, and the generation of non-consensual explicit images of minors, manageable by law enforcement.