Purportedly AI-Generated Images Allegedly Linked Sami Lushtaku to Banjska Attack Suspect Milan Radoičić

The Facebook page ZËRI Shqiptar reportedly published AI-generated images purporting to show Sami Lushtaku, the mayor of Skenderaj, Kosovo, traveling in a luxury car with Milan Radoičić, a former Serb List official accused in connection with the 2023 Banjska attack on Kosovo police. The reportedly AI-generated images appeared amid a political dispute after Vetëvendosje officials accused Lushtaku of assaulting Hysni Mehani, the party's local leader in Skenderaj.

Risk classification

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

The incident involved the creation of a deepfake using a public figure's likeness to target and politically manipulate public perception during a local dispute.

Additional risk subdomains

• 3.1 False or misleading information: The AI system was used to generate false photographic evidence to mislead the public about a politician's associations.

Causal factors

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

The incident was caused by human actors who intentionally generated and distributed a political deepfake post-deployment to mislead the public.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The report describes AI-generated content such as deepfakes, which fall under Risk Level 3 due to transparency and disclosure requirements.

AI system and alleged parties

• AI system: HiveModeration, unspecified (Hive)
• AI purpose: Deepfake Image Generation; Image Verification
• Behaviour type: Tool
• Alleged developer: Synthetic media generation technology developers, Image generation technology developers, Deepfake technology developers
• Alleged deployer: ZËRI Shqiptar (Facebook), Information manipulation actors in Kosovo, Information manipulation actors
• Alleged harmed parties: Voters in Kosovo, Voters, Sami Lushtaku, Epistemic integrity, Democratic integrity

Harm severity

Severity ratings are based on this scale.

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

Malicious content

Reported: Yes, the report describes the creation and dissemination of AI-generated fake images on Facebook.

Directly caused: An AI generator was used to create fabricated images of the mayor of Skenderaj traveling with a suspected terrorist, which were shared on Facebook.

Inferred additional harm: It is likely that thousands of Facebook users were exposed to this malicious content before it was debunked.

Epistemic

Reported: Yes, the report describes the fabrication of fake photographic evidence to mislead the public.

Directly caused: AI-generated fake photos were created and circulated on Facebook to falsely show a political figure traveling with a terrorist suspect.

Inferred additional harm: The incident likely contributed to public confusion and eroded trust in photographic evidence during a politically sensitive period.

People affected

• Occurrences reported: 1
• People reportedly harmed: 1
• People reportedly exposed: 1000

Potential causes

Management

• Insufficient Moderation Budget: Platforms allocated inadequate resources for monitoring regional content.

Technology

• Generative AI Accessibility: Easy access to AI tools allowed actors to create highly realistic fake images.
• Lack of Platform AI Detection: Facebook lacked automated tools to instantly flag AI-generated content.

Data Inputs

• Availability of Source Photos: Publicly available photos of politicians were used to train the generator.

Human Factors

• Intentional Political Deception: Malicious actors created the deepfake to exploit existing local tensions.
• Uncritical Sharing by Public: Social media users shared the image without verifying its authenticity.

Process and Methods

• Reactive Fact-Checking Model: Fact-checking occurred only after the deepfake had already gone viral.

Regulatory Environment

• Weak Synthetic Media Laws: Lack of legal frameworks to deter and punish political deepfake creation.

Information quality

• Classification confidence: High
• Reason for confidence: The fact-checking report clearly details the nature of the fake image, the political context, and the verification method using HiveModeration.
• Ambiguities identified: The specific AI model used to generate the deepfake is not identified in the report.

An AI-generated deepfake was used in Kosovo to falsely depict a local mayor traveling with a suspected terrorist during a period of political tension. The incident represents a minor national security concern, characterized as a localized information operation that was quickly debunked by fact-checkers, with negligible impact on broader state sovereignty or physical security.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: Kosovo
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. The specific deepfake has been debunked, representing an ongoing strategic concern regarding political disinformation rather than an active, immediate national crisis.
• Autonomy: Human-controlled. The AI system was used as a basic tool by human actors to generate synthetic media based on specific user prompts.
• Novelty: Established threat. The creation and dissemination of political deepfakes to damage reputations is a well-established threat vector globally.

Impact by dimension

• Physical security: Negligible. The incident involved a digital image shared on social media with no physical impact, kinetic threats, or critical infrastructure compromise.
• Information security: Minor. AI-generated deepfake used in a localized political disinformation campaign. It was quickly identified and debunked by fact-checkers, limiting its intelligence impact.
• Sovereignty: Minor. The deepfake targeted a municipal mayor during local political tensions, representing a minor attempt to manipulate public perception of local government figures.
• Economic security: Negligible. No financial systems, strategic technologies, or economic supply chains were affected or targeted in this incident.
• Societal stability: Minor. The incident contributed to epistemic confusion and local political polarization, but did not lead to widespread civil unrest or systemic human rights abuses.