Bangladesh-Operated Facebook Pages Reportedly Shared AI-Generated Anti-Data-Center Imagery

U.S.-state-themed Facebook pages reportedly operated from Bangladesh shared purportedly AI-generated imagery opposing data-center development. The pages allegedly used local-sounding names to push hyperlocal grievance content into U.S. communities, making offshore engagement farming appear like grassroots opposition.

Risk classification

• Primary risk domain: 4 Malicious actors
• Primary risk subdomain: 4.1 Disinformation, surveillance, and influence at scale

The incident involves coordinated, at-scale foreign influence and disinformation campaigns using AI-generated content to manipulate public opinion and political processes in the U.S.

Additional risk subdomains

• 3.1 False or misleading information: The campaigns generated specific false stories, such as fake earthquakes in various states and fabricated tech news articles.

Causal factors

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

The risk is caused by the intentional decisions and actions of human adversaries (foreign state actors and spam operators) using deployed AI systems to conduct influence operations.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The report describes the use of chatbots (ChatGPT) and AI-generated content (images, videos, and articles) to influence public opinion, which falls under the category of AI-generated content requiring transparency obligations.

AI system and alleged parties

• AI system: ChatGPT, unspecified (OpenAI)
• AI purpose: Social Media Content Generation; Image Generation
• Behaviour type: Tool
• Alleged developer: Synthetic media generation technology developers, Image generation technology developers
• Alleged deployer: Operators of inauthentic social media accounts, Operators of AI-generated content farms, Information manipulation actors in Bangladesh, Information manipulation actors
• Alleged harmed parties: Threads users, Social media users, Public AI infrastructure, People participating in U.S. AI policy debates, Meta users, Instagram users, Facebook users, Epistemic integrity, Data center developers, Communities debating AI data center development

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 Minor, indirect Minor
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Minor, indirect Substantial
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Malicious content

Reported: Yes, the report explicitly describes the creation and spread of malicious and deceptive content.

Directly caused: The campaigns directly generated AI 'slop', including obscene hand gestures carved into crops, fake comic strips, and AI-narrated 'breaking news' videos designed to stoke outrage.

Inferred additional harm: It is highly likely that millions of additional AI-generated spam images and videos were distributed across Meta and TikTok platforms, reaching a wider audience than documented.

Democracy

Reported: Yes, the report explicitly describes the erosion of democratic norms through foreign interference.

Directly caused: Foreign state actors from China, Russia, and Iran used AI-generated content to exploit domestic political divisions and influence public opinion ahead of U.S. elections.

Indirectly caused: The campaigns aim to weaken trust in U.S. infrastructure, companies, and government institutions by turning local grievances into national wedge issues.

Inferred additional harm: Continued exposure to coordinated foreign influence operations could further polarize the electorate and undermine faith in democratic processes and public institutions.

Epistemic

Reported: Yes, the report explicitly describes epistemic harm through the spread of fabricated information.

Directly caused: The networks fabricated news stories, including fake TechCrunch and Gizmodo articles about data center vulnerabilities, and posted false claims of widespread earthquakes with inaccurate maps.

Indirectly caused: The high volume of AI-generated 'slop' and coordinated inauthentic accounts pollutes the information ecosystem, making it harder for citizens to find reliable information.

Inferred additional harm: The proliferation of highly realistic fake news and AI-generated content is likely to cause long-term degradation of shared consensus reality among targeted social media users.

People affected

• Occurrences reported: 1
• People reportedly exposed: 10000

Potential causes

Management

• Underfunded Trust and Safety: AI developers and social platforms underinvest in abuse prevention.
• Prioritizing Engagement Over Truth: Platforms focus on user engagement, which favors sensational AI slop.
• Inadequate Risk Assessment: Tech firms failed to anticipate LLM exploitation for infrastructure sabotage.

Technology

• Generative AI Abuse: ChatGPT used by foreign actors to create convincing fake articles and comics.
• AI Image Generators: Used to mass-produce anti-data center slop and localized outrage memes.
• Algorithmic Amplification: Social media algorithms prioritize high-engagement AI-generated content.

Data Inputs

• Inauthentic Prompting: Operators feed manipulative prompts to generate targeted political propaganda.
• Lack of Provenance Data: Absence of metadata to verify if social media images are AI-generated.
• Misleading Localized Inputs: Prompts tailored with specific state names to mimic local grassroots voices.

Human Factors

• Cognitive Vulnerability: Public susceptibility to emotionally charged, hyper-local AI-generated bait.
• Domestic Political Exploitation: Local political figures adopting adversary AI aesthetics for campaign gains.
• Credulity Toward Fake Media: Users trust fake TechCrunch/Gizmodo articles written by generative AI.

Process and Methods

• Inadequate Bot Detection: Platforms fail to quickly detect and block AI-driven inauthentic networks.
• Poor Content Provenance Tracking: Social networks lack systems to flag synthetic content at scale.
• Delayed Attribution Processes: Threat intelligence identifies AI campaigns only after they achieve reach.

Regulatory Environment

• Lack of Synthetic Media Laws: No strong regulatory framework to govern AI-generated disinformation.
• Weak Infrastructure Protection: Regulatory gaps in protecting critical AI compute facilities from propaganda.
• Unregulated Social Platforms: Insufficient legal pressure on platforms to clean up automated AI slop.

Information quality

• Classification confidence: High
• Reason for confidence: The reports from the New York Times, Alethea, and Conspirator0 provide highly detailed, consistent, and cross-verified accounts of the foreign influence campaigns, the platforms used, the specific themes of the AI-generated content, and the actors involved (Bangladesh spam network, Storm-1516, Rybar, and Chinese state media).
• Ambiguities identified: The exact reach and measurable impact of these campaigns on public opinion remain difficult to quantify, as noted by OpenAI finding 'little to no authentic engagement' for some campaigns.
• Alternative interpretations: Some actors argue the opposition to data centers is entirely organic, though threat intelligence platforms documented clear coordination and inorganic elements.

Foreign state and non-state actors from China, Russia, Iran, and Bangladesh deployed generative AI tools to conduct coordinated influence operations targeting U.S. critical data infrastructure and democratic processes. While direct engagement was limited, the incident highlights the evolving threat of AI-scaled disinformation designed to exploit domestic political divisions.

• Overall national security impact: Substantial
• Scope: Multiple nations
• Primary target: United States
• Other affected: Armenia
• Alleged perpetrator: China, Russia, Iran, and Bangladesh-based networks

Threat characteristics

• Imminence: Long-term. Represents an ongoing strategic information warfare capability and influence concern rather than an immediate, kinetic crisis.
• Autonomy: Human-controlled. AI tools were directly prompted and deployed by human threat actors to generate specific text and image assets.
• Novelty: Evolved capability. Represents a significant scaling and sophistication of traditional foreign influence operations through the integration of generative AI tools.

Impact by dimension

• Physical security: Negligible. No direct physical attacks, kinetic actions, or compromise of critical infrastructure control systems were reported.
• Information security: Substantial. Foreign state actors from China, Russia, and Iran conducted coordinated, AI-enhanced disinformation campaigns using fake articles and synthetic media to influence US public opinion.
• Sovereignty: Minor. The campaigns targeted democratic discourse and public trust ahead of US midterm elections, though they achieved limited authentic engagement.
• Economic security: Minor. The operations aimed to undermine public support for critical U.S. data center infrastructure, potentially impacting technological development and AI computational expansion.
• Societal stability: Minor. The campaigns utilized rural rage bait and localized grievances to deepen political polarization, though no civil unrest or mass instability resulted.