Female Politicians in the United Kingdom Reportedly Victimized by Purported Deepfake Pornography

Female politicians in the United Kingdom, including Angela Rayner, Gillian Keegan, Penny Mordaunt, Priti Patel, Stella Creasy, and Dehenna Davison, have reportedly been targeted by nonconsensual, purportedly AI-generated deepfake pornography. The images, which some accounts suggest may have circulated online for an extended period, were reported to have caused distress and, in at least some instances, to have prompted reports to or involvement by law enforcement.

Risk classification

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

The incident primarily involves the creation of sexually explicit deepfakes to target, humiliate, and manipulate specific female politicians, which falls under targeted manipulation and humiliating imagery.

Additional risk subdomains

• 4.1 Disinformation, surveillance, and influence at scale: The reports also detail AI-generated political deepfakes designed to mislead voters, disrupt democratic elections, and manipulate public opinion at scale.

Causal factors

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

The incidents were caused by human actors intentionally using deployed AI tools to generate and distribute nonconsensual pornography and political disinformation.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Risk Level 3: Limited Risk. The report describes the creation of 'deepfakes' and 'AI-generated' content, which are subject to transparency obligations under the EU AI Act.

AI system and alleged parties

• AI system: ClothOff, unspecified
• AI purpose: Deepfake Video Generation; Deepfake Image Generation
• Behaviour type: Tool
• Alleged developer: Unknown image generator developers, Unknown deepfake technology developers
• Alleged deployer: Unknown deepfake creators
• Alleged harmed parties: Women, Stella Creasy, Priti Patel, Penny Mordaunt, Gillian Keegan, Female politicians in the United Kingdom, Female politicians, Epistemic integrity, Dehenna Davison, Angela Rayner

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

Malicious content

Reported: Yes

Directly caused: AI-generated sexually explicit videos and images of female politicians were uploaded to pornographic websites, attracting hundreds of thousands of views.

Inferred additional harm: Millions of sexually explicit deepfakes are estimated to circulate online, exposing a vast audience to nonconsensual toxic content.

Differential treatment

Reported: Yes

Directly caused: The report notes that 99% of sexually explicit deepfakes target females, representing a highly gendered form of digital violence and harassment.

Inferred additional harm: This gendered targeting likely discourages women from entering public life or politics, exacerbating gender inequality.

Civil rights

Reported: Yes

Directly caused: The nonconsensual sexualization of women's bodies violates their basic rights to bodily autonomy, digital dignity, and privacy.

Inferred additional harm: The widespread use of deepfakes threatens the civil rights of individuals to control their own likeness and protect themselves from targeted harassment.

Democracy

Reported: Yes

Directly caused: Deepfakes misrepresenting politicians, such as George Freeman's fake defection or the fake withdrawal of Irish candidate Connolly, directly undermine democratic processes.

Indirectly caused: The first round of the Romanian presidential elections in November 2024 was annulled following credible reports of Russian interference involving deepfake videos.

Inferred additional harm: Widespread political deepfakes risk eroding public trust in democratic institutions, leading to voter apathy and the potential subversion of election outcomes.

Privacy

Reported: Yes

Directly caused: The faces and likenesses of politicians were used without their consent to generate explicit pornographic images and misleading political videos.

Inferred additional harm: The ease of creating deepfakes threatens the privacy of any individual whose public photos can be weaponized against them.

Psychological

Reported: Yes

Directly caused: Politicians like Dehenna Davison felt 'quite violating', Stella Creasy felt 'sick', and US Representative experienced resurfaced trauma from encountering a deepfake of herself.

Inferred additional harm: Widespread psychological distress, anxiety, and trauma are highly likely among the thousands of unnamed victims of nonconsensual deepfake pornography.

Epistemic

Reported: Yes

Directly caused: A fake video of George Freeman defecting to Reform UK and a bogus news report of a coup in France misled viewers and public officials.

Inferred additional harm: The proliferation of highly convincing deepfakes feeds a general sense of epistemic skepticism where citizens feel they cannot believe anything they see or hear.

People affected

• Occurrences reported: 35
• People reportedly harmed: 38
• People reportedly exposed: 38

Potential causes

Management

• Lack of Platform Accountability: Tech companies prioritize hosting content over verifying its authenticity.
• Inadequate Risk Assessment by Apps: AI apps like ClothOff fail to implement safeguards against abusive use.

Technology

• Advanced Deepfake Generation Tools: AI tools can easily strip clothes or generate highly realistic fake videos.
• Low Barrier to Entry for AI: Widespread availability of AI technology enables anyone to create deepfakes.
• Sophisticated Digital Hallucinations: AI fakes are so realistic that viewers can no longer tell they are false.

Data Inputs

• Unauthorized Use of Public Images: AI models use public photos of politicians to generate explicit deepfakes.
• Lack of Input Data Verification: Open-source images are ingested by AI apps without consent or verification.

Human Factors

• Malicious Intent of Creators: Users weaponize AI to humiliate, disempower, and violate women's bodies.
• Warped Understanding of Consent: Exposure to deepfake porn warps sexual understanding of consent in youth.
• Anonymity of Online Perpetrators: Creators operate from the comfort of anonymity with no fear of comeback.

Process and Methods

• Social Media Monitoring Gaps: Platforms lack proactive mechanisms to detect and block AI-generated fakes.
• Inadequate Platform Moderation: Meta and other platforms refuse to remove deepfakes, citing no policy breach.
• Easy Search Engine Accessibility: Mainstream search engines easily index and direct users to deepfake sites.

Regulatory Environment

• Legal Loophole for Image Creation: Creating deepfakes remains legal in the UK despite distribution bans.
• Absence of Identity Theft Laws: Police have no legal basis to act against non-consensual identity fakes.
• Delayed Legislative Action: Proposed laws to ban deepfake creation were dropped due to early elections.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide clear, detailed, and consistent accounts of both deepfake pornography targeting female politicians and political deepfakes targeting various candidates. The sources are reputable and corroborate each other on the societal and political impacts of these technologies.
• Ambiguities identified: The exact technical models used to generate the political deepfakes (other than the mention of ClothOff for some pornographic deepfakes) are not specified.
• Alternative interpretations: None. The events clearly represent malicious use of AI for harassment and political manipulation.

The widespread generation of nonconsensual deepfake pornography targeting female politicians and misleading political videos in the UK represents a substantial threat to democratic integrity, information security, and societal stability. The incidents have prompted direct intervention by the UK Electoral Commission, highlighting the growing national security challenge of AI-enabled political manipulation.

• Overall national security impact: Substantial
• Scope: Multiple nations
• Primary target: United Kingdom
• Other affected: Romania, Moldova, Ireland, France
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Near-term. Ongoing and active threat to upcoming elections and political figures, requiring continuous monitoring and mitigation.
• Autonomy: Human-controlled. AI systems function as tools used intentionally by human creators to generate explicit or misleading media.
• Novelty: Evolved capability. While deepfakes are an established technology, the scale and ease of targeting political figures represents an evolved capability.

Impact by dimension

• Physical security: Negligible. No physical threats, kinetic attacks, or critical infrastructure impacts reported in relation to these deepfake incidents.
• Information security: Substantial. AI-generated disinformation and deepfakes targeted UK political figures to misrepresent stances and manipulate public perception, prompting a national electoral security response.
• Sovereignty: Substantial. Deepfakes misrepresenting political figures and false defections directly threaten the integrity of democratic processes and electoral trust.
• Economic security: Negligible. No significant threats to financial systems, strategic technology theft, or economic stability reported in this incident.
• Societal stability: Substantial. Widespread nonconsensual deepfake pornography targeting female politicians violates human rights, privacy, and represents highly gendered digital harassment.