Fake LinkedIn Profiles Created Using GAN Photos

More than a thousand inauthentic LinkedIn profiles using allegedly GAN-generated photos were notified by researchers at Stanford to LinkedIn’s staff, and many of which were removed for violating rules against creating fake profiles and falsifying information.

Risk classification

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

The incident involved using AI-generated synthetic faces to create fake professional personas on LinkedIn to manipulate and deceive users into engaging with automated sales pitches for commercial advantage.

Causal factors

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

The deployment of AI-generated profile pictures to create fake accounts was an intentional decision by human actors (marketing vendors) to conduct automated sales outreach.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The report describes the use of AI-generated content (deepfakes/synthetic profile pictures), which under the EU AI Act falls under limited risk and requires transparency obligations to inform users they are interacting with AI-generated content.

AI system and alleged parties

• AI system: Generative Adversarial Network
• AI purpose: Deepfake Image Generation
• Behaviour type: Tool
• Alleged developer: unknown
• Alleged deployer: unknown
• Alleged harmed parties: LinkedIn users

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 Negligible, 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 Minor
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Epistemic

Reported: The report explicitly describes epistemic harm through the creation of over 1,000 fake profiles using AI-generated faces, which eroded the expectation of authenticity and truth on professional social networks.

Directly caused: Over 1,000 fake LinkedIn profiles were created using synthetic GAN-generated faces, misleading users into believing they were interacting with real professionals.

Indirectly caused: The proliferation of these profiles contributed to a broader erosion of trust on social media platforms, making users more skeptical of online interactions.

Inferred additional harm: It is likely that many more users than the reported researchers were misled by these profiles, potentially altering their professional networking behavior due to decreased trust.

People affected

• Occurrences reported: 1
• People reportedly exposed: 1000

Potential causes

Management

• Outsourcing Lead Generation: Companies outsourced sales outreach to third parties without oversight.
• Prioritizing Low-Cost Outreach: Using AI avatars is cheaper than hiring real sales staff to scale operations.

Technology

• High-Fidelity GAN Generation: GANs generate highly realistic faces that are difficult to distinguish.
• Easy Access to AI Tools: Websites allow anyone to download computer-generated faces for free.

Data Inputs

• Large Training Datasets: GANs trained on huge photo datasets to produce lifelike human features.

Human Factors

• Limited Human Detection Ability: Humans have only a 50 percent chance of identifying AI-generated faces.
• Perceived Trustworthiness of AI: AI-synthesized faces are rated as more trustworthy than real human faces.
• Expectation of Authenticity: Users assume profiles represent real humans on professional networks.

Process and Methods

• Bypassing Platform Limits: Fake profiles used to circumvent LinkedIn's messaging limits.
• Inadequate Vendor Auditing: Clients failed to audit lead-generation methods used by outside vendors.
• Automated Detection Gaps: LinkedIn's automated defenses failed to catch all GAN-generated profiles.

Regulatory Environment

• Ambiguous Avatar Rules: Some vendor CEOs believed there were no specific rules against AI avatars.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide consistent, detailed accounts of the Stanford Internet Observatory's investigation and NPR's follow-up verification. The facts regarding the scale (over 1,000 profiles), the technology used (GANs), and the commercial motivation (lead generation) are clearly documented across multiple sources.
• Ambiguities identified: The exact identity of all third-party vendors who created the profiles remains partially unverified, as some companies declined to comment or scrubbed their websites.
• Alternative interpretations: None, the commercial lead-generation context is clearly established as the primary driver of the incident.

The incident involves the deployment of over 1,000 fake LinkedIn profiles using GAN-generated headshots for automated commercial sales outreach. While representing an evolved capability in scaling synthetic identities, the incident was purely commercial, lacked state-actor involvement, and was successfully resolved through standard platform moderation, posing minimal threat to national security.

• Overall national security impact: Minor
• Scope: Multiple nations
• Primary target: No clear primary
• Other affected: Multiple nations
• Alleged perpetrator: Third-party lead-generation vendors

Threat characteristics

• Imminence: Long-term. Represents an ongoing strategic concern regarding identity authentication rather than an active national security crisis.
• Autonomy: Human-supervised. AI generated the profile images, but human operators created the accounts, set up the automation parameters, and managed the sales outreach.
• Novelty: Evolved capability. GAN-generated faces are an established technology, but using them to scale over 1,000 fake personas for automated commercial outreach represents an evolved deployment capability.

Impact by dimension

• Physical security: Negligible. No physical threat, kinetic targeting, or critical infrastructure compromise occurred in this commercial sales incident.
• Information security: Minor. Use of over 1,000 GAN-generated profiles for commercial outreach. While not a state-sponsored campaign, it demonstrates the scalability of synthetic identities, resolved via standard platform moderation.
• Sovereignty: Negligible. No impact on government decision-making, electoral systems, or sovereign state authority.
• Economic security: Negligible. No threat to strategic industries, critical supply chains, or financial systems; limited to unauthorized commercial B2B sales outreach.
• Societal stability: Negligible. No mass surveillance, systemic human rights violations, or large-scale societal instability resulted from this sales-focused campaign.