Noelle Martin Deepfaked Without Consent in AI-Generated Pornography

In 2017, Noelle Martin discovered explicit deepfake videos online that used AI technology to superimpose her face onto pornographic scenes. This incident was a continuation of the abuse she had experienced since at least 2012, when she first found doctored still images of herself in similar contexts. Despite the initial lack of legal protections, her advocacy efforts were instrumental in making image-based abuse a criminal offense in Australia.

Risk classification

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

The incident involved the creation of humiliating and non-consensual sexual imagery targeting a specific victim to harass and degrade her.

Additional risk subdomains

• 1.2 Exposure to toxic content: The victim and online users were exposed to highly toxic, non-consensual pornographic deepfake videos generated using AI.

Causal factors

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

The harm was caused by human actors intentionally using deployed AI deepfake generation tools to target the victim with non-consensual pornographic videos.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The report describes AI-generated content such as deepfakes, which are classified under Limited Risk and subject to transparency obligations.

AI system and alleged parties

• AI system: Face2Face, FaceApp, Zao, unspecified
• AI purpose: Deepfake Video Generation; Deepfake Image Generation
• Behaviour type: Tool
• Alleged developer: Zao, University of Erlangen-Nuremberg, Stanford University, Max Planck Institute, FaceApp, Face2Face
• Alleged deployer: Unknown deepfake creators
• Alleged harmed parties: Noelle Martin

Harm severity

Highest direct severity in any category: Severe. 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 Substantial
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Minor, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Minor, indirect Negligible
• Psychological: direct Minor, indirect Substantial
• Epistemic: direct Minor, indirect Substantial
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Malicious content

Reported: Yes

Directly caused: Perpetrators created and uploaded highly realistic 11-second deepfake pornographic videos of the victim to adult websites.

Indirectly caused: The technology enabled the viral spread of celebrity deepfake porn on platforms like Reddit and Twitter before bans were implemented.

Inferred additional harm: Millions of internet users were potentially exposed to these toxic, non-consensual sexual videos across various adult hosting sites.

Civil rights

Reported: Yes

Directly caused: The victim's right to control her own image and bodily representation was severely violated by the creation of non-consensual sexual deepfakes.

Indirectly caused: N/A

Inferred additional harm: The widespread availability of deepfake tools threatens the fundamental rights to privacy, dignity, and personal security for women globally.

Privacy

Reported: Yes

Directly caused: Perpetrators used the victim's public Facebook photos without her consent to train AI models and generate deepfake pornographic videos.

Indirectly caused: N/A

Inferred additional harm: Thousands of individuals have had their personal images harvested from social media to create non-consensual synthetic content, violating their privacy on a massive scale.

Psychological

Reported: Yes

Directly caused: The victim experienced panic, horror, dryness of mouth, heavy heart thudding, numbness, and deep pain upon discovering the deepfake videos of herself.

Indirectly caused: The victim suffered long-term emotional distress, isolation, and shame, feeling forced to stand alone in her fight for nearly a decade.

Inferred additional harm: Other targeted individuals, including the named celebrities and unnamed victims, likely suffered similar severe psychological trauma, anxiety, and distress.

Epistemic

Reported: Yes

Directly caused: The highly convincing deepfake videos made the victim worry that anyone seeing them would believe they were real, eroding the boundary between truth and fabrication.

Indirectly caused: The rise of user-friendly deepfake apps contributes to a broader societal shift where video is no longer trusted as proof of reality.

Inferred additional harm: Widespread deepfake technology likely causes systemic epistemic harm by making it difficult for the public to distinguish real footage from synthetic fabrications.

People affected

• Occurrences reported: 1
• People reportedly harmed: 4
• People reportedly exposed: 4

Potential causes

Management

• Delayed Platform Bans: Reddit and Twitter delayed banning explicit deepfakes until 2018.
• Prioritizing Growth over Safety: Platforms prioritized user growth over implementing robust safety safeguards.
• Lack of Proactive Risk Assessment: Tech companies failed to foresee the weaponization of deep learning tech.

Technology

• Accessible Deep Learning Software: Open-source and user-friendly apps allowed anyone to generate deepfakes.
• Realistic Facial Reanimation: Systems like Face2Face enabled highly convincing facial manipulation.
• Rapid AI Tech Proliferation: The rapid spread of deepfake algorithms outpaced safety barriers.

Data Inputs

• Unprotected Social Media Photos: Perpetrators easily scraped personal photos uploaded to Facebook.
• Abundance of Training Data: Large volumes of online images allowed AI to learn and replicate faces.
• Lack of Image Provenance: No metadata or watermarks prevented unauthorized reuse of personal images.

Human Factors

• Malicious Intent of Creators: Perpetrators targeted individuals for harassment and non-consensual porn.
• Inherent Trust in Video Content: Audiences naturally trust video evidence, magnifying the impact of fakes.
• Social Media Over-sharing: Users posted personal photos without anticipating malicious AI exploitation.

Process and Methods

• Weak Platform Moderation: Websites and platforms failed to proactively detect and block deepfakes.
• Ineffective Takedown Processes: Host sites ignored requests or allowed deleted content to be re-uploaded.
• Lack of Host Accountability: Some site administrators extorted victims instead of removing content.

Regulatory Environment

• Lagging Cyber Abuse Laws: Existing laws in 2012 lacked provisions for non-consensual image abuse.
• Lack of Global Enforcement: Jurisdictional boundaries made it difficult to police international hosts.
• Inadequate Platform Penalties: No regulatory penalties existed to force platforms to remove fake content.

Information quality

• Classification confidence: High
• Reason for confidence: The report is a detailed, first-person account of the victim's experience with deepfake technology, providing clear timelines, emotional impacts, and specific technological context (Face2Face, FaceApp, Zao). The facts regarding the AI's role as a tool for generating non-consensual sexual imagery are unambiguous.
• Ambiguities identified: None of significance regarding the AI safety incident itself, though the exact technical algorithms used by the specific perpetrators to target Noelle Martin are not named.
• Alternative interpretations: None. The incident is a clear-cut case of targeted harassment and non-consensual deepfake pornography.

The incident involves the targeted creation of non-consensual AI-generated deepfake pornography against an Australian citizen and global celebrities. While causing severe psychological harm and highlighting regulatory gaps in managing synthetic media, its direct implications for national security remain minor, manageable through standard legal and domestic policy frameworks.

• Overall national security impact: Minor
• Response level: Moderate
• Scope: Multiple nations
• Primary target: No clear primary
• Other affected: Australia, United States
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. Represents an ongoing societal and legal challenge regarding synthetic media rather than an active, imminent national security crisis.
• Autonomy: Human-controlled. Perpetrators actively selected targets and used AI tools to generate and upload the synthetic media.
• Novelty: Evolved capability. Represents a significant advancement of image manipulation from manual photoshopping to automated, realistic AI-generated deepfakes.

Impact by dimension

• Physical security: Negligible. No physical systems, critical infrastructure, or kinetic targeting systems were affected in this incident.
• Information security: Negligible. While involving synthetic media, the incident was a targeted harassment campaign against individuals rather than a state-sponsored information warfare or intelligence operation.
• Sovereignty: Negligible. No impact on state authority, government decision-making, or sovereign operations.
• Economic security: Negligible. No strategic economic assets, financial systems, or critical technology supply chains were compromised.
• Societal stability: Minor. Represents a notable violation of personal privacy and human rights, which prompted legislative changes and the criminalization of image-based abuse in Australia.