Social Media Algorithms Amplified Disinformation Campaign in Honduras Election

In October 2021, a coordinated network of over 317 fake Twitter accounts leveraged AI-driven algorithms to amplify disinformation about the Honduran presidential election, targeting opposition candidate Xiomara Castro. The campaign spread false narratives to suppress voter turnout and undermine the election's integrity. Social media platforms, including Twitter and Facebook, removed the accounts only after being alerted, which also raised concerns about inadequate moderation.

Risk classification

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

The incident involved a coordinated, at-scale online influence campaign using hundreds of fake accounts to spread political disinformation and manipulate the Honduran democratic election.

Additional risk subdomains

• 3.2 Pollution of information ecosystem and loss of consensus reality: The coordinated campaign created a fake news feedback loop that flooded the political discourse and sowed distrust among voters.
• 6.5 Governance failure: Social media platforms failed to adequately moderate and prevent repeat political disinformation campaigns in developing countries like Honduras.

Causal factors

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

The disinformation campaign was intentionally designed and deployed by human political actors post-deployment to manipulate the Honduran election.

EU AI Act risk tier

• Risk tier: 1 Unacceptable

Unacceptable Risk: The report describes 'AI used for subliminal manipulation or behavioral influence leading to harm', specifically a coordinated campaign to manipulate political beliefs and deter citizens from voting, which violates fundamental rights and democratic processes.

AI system and alleged parties

• AI system: AI-driven algorithms
• AI purpose: Social Media Content Generation; Content Curation
• Behaviour type: Tool
• Alleged developer: X (Twitter), Meta, Facebook
• Alleged deployer: Unknown Twitter users, Unknown Facebook users, National Party of Honduras supporters, Juan Orlando Hernández supporters
• Alleged harmed parties: Xiomara Castro, Libertad y Refundación (LIBRE) supporters, Honduras, Honduran electorate, Electoral integrity, Democracy

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

Malicious content

Reported: The report explicitly describes malicious content created and spread directly by the coordinated network of fake accounts.

Directly caused: The network of at least 317 fake Twitter accounts spread conspiracy theories, false claims about political alliances, and content about alleged corruption to discourage voting.

Indirectly caused: A fake news website, La Tribuna Honduras, amplified the disinformation by reporting on the fake Twitter activity, creating a fake news feedback loop.

Inferred additional harm: It is likely that similar malicious content was spread across other social media platforms like Facebook, where the ruling party's page previously received fake engagement.

Civil rights

Reported: The report explicitly describes attempts to interfere with civil rights, specifically the right of Honduran citizens to vote.

Directly caused: The coordinated campaign directly targeted Honduran voters with messages discouraging them from voting, saying 'I won't even go to vote' to suppress turnout.

Indirectly caused: The manipulation of the information environment undermined the integrity of the democratic process and the citizens' ability to make free electoral choices.

Inferred additional harm: The suppression of votes may have altered the election outcome, further impacting the civil and political rights of the Honduran population under the resulting administration.

Democracy

Reported: The report explicitly describes the erosion of democratic norms caused directly and indirectly by the incident.

Directly caused: The coordinated campaign aimed to affect the integrity of the democratic process by spreading conspiracies about opposition candidates and deterring citizens from voting.

Indirectly caused: The campaign sowed distrust and confusion among voters, weakening public trust in democratic institutions and elections.

Inferred additional harm: The ongoing use of such digital manipulation tactics by ruling parties in Latin America threatens the long-term stability of democratic governance in the region.

Privacy

Reported: The report explicitly describes privacy violations where photos of unsuspecting individuals were used without consent.

Directly caused: The profile photos on the fake accounts linked back to the Facebook pages of unsuspecting Peruvians, using their likenesses as avatars without their knowledge or consent.

Inferred additional harm: It is likely that hundreds of other individuals had their public social media photos and identities compromised to populate these fake networks.

Epistemic

Reported: The report explicitly describes epistemic harm through the generation and amplification of false political narratives.

Directly caused: The network of fake accounts falsely suggested that candidate [PERSON_005] might join forces with a convicted felon, Yani Rosenthal, to mislead voters.

Indirectly caused: The fake news feedback loop created by La Tribuna Honduras reporting on the fake tweets as real voter sentiment further corrupted the shared information ecosystem.

Inferred additional harm: The widespread dissemination of fake news likely led to a persistent loss of consensus reality among the Honduran electorate regarding the candidates' actual platforms and alliances.

People affected

• Occurrences reported: 1
• People reportedly harmed: 2
• People reportedly exposed: 317

Potential causes

Management

• Disproportionate Safety Spending: Platforms spent 87 percent of misinformation budget on English markets.
• Prioritization of Profit Over Safety: Facebook invested more in high-profit regions, neglecting developing nations.
• Delayed Response to Warnings: Management took nearly a year to act on internal whistleblower warnings.

Technology

• Algorithmic Amplification of Fake News: Algorithms amplified coordinated inauthentic posts and fake news articles.
• Inadequate Automated Detection: Automated systems failed to proactively block coordinated bot networks.

Data Inputs

• Farmed and Compromised Accounts: Decade-old compromised accounts were easily purchased and repurposed.
• Stolen Profile Photos: Accounts used photos of real, unsuspecting Peruvians to bypass basic checks.

Human Factors

• Voter Susceptibility to Fake News: Voters believed coordinated disinformation due to a lack of verification.
• Exploitation by Political Actors: Ruling party staff and PR firms actively manipulated social media platforms.

Process and Methods

• Reactive Moderation Process: Platforms acted only after external cybersecurity firms reported the network.
• Ineffective Enforcement Loop: Banned actors easily returned because platforms only gave slaps on the wrist.

Regulatory Environment

• Lack of Local Platform Regulation: Developing nations lack legal framework to penalize foreign tech companies.
• Weak Election Integrity Oversight: Local election bodies failed to prevent digital manipulation of campaigns.

Information quality

• Classification confidence: High
• Reason for confidence: The report provides clear, detailed evidence from cybersecurity firm Nisos and former platform employees regarding the scale, tactics, and targets of the coordinated disinformation campaign. The role of automation in the campaign is well-documented, and the political context is explicitly described.
• Ambiguities identified: The exact technical complexity of the automation tools (whether they used advanced generative AI or simple database-driven scripts) is not fully specified.
• Alternative interpretations: The campaign could be viewed purely as a human-driven political PR effort rather than an AI safety incident, though the use of automated bot networks to manipulate discourse qualifies it as an algorithmic safety risk.

A coordinated network of 317 fake accounts used automated tools to spread political disinformation and suppress voter turnout during the 2021 Honduran presidential election. While the technical sophistication was low and the immediate threat has been mitigated, the incident highlights the ongoing vulnerability of democratic processes to automated foreign or domestic influence operations.

• Overall national security impact: Substantial
• Scope: Single nation
• Primary target: Honduras
• Other affected: Peru
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. The specific campaign has been resolved by platform intervention, but it represents an ongoing strategic threat of digital election manipulation.
• Autonomy: Human-controlled. The automated accounts performed pre-programmed functions of posting identical tweets simultaneously, acting as basic tools under human direction.
• Novelty: Established threat. Coordinated inauthentic behavior and bot-driven disinformation campaigns are well-established tactics previously observed globally.

Impact by dimension

• Physical security: Negligible. No physical systems, infrastructure, or human safety threats were reported in connection with this digital disinformation campaign.
• Information security: Substantial. The incident involved a coordinated network of 317 fake accounts using automated techniques to spread political disinformation and create fake news feedback loops.
• Sovereignty: Substantial. The campaign targeted a national presidential election, attempting to manipulate the democratic process and suppress voter turnout against opposition candidates.
• Economic security: Negligible. No strategic technology theft, financial system attacks, or economic manipulation were indicated in the incident details.
• Societal stability: Substantial. The operation targeted the Honduran electorate at scale to suppress votes, undermining democratic norms, and violated the privacy of unsuspecting individuals whose photos were stolen.