This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A coordinated network of over 317 fake Twitter accounts used automated techniques to amplify disinformation during the 2021 Honduran presidential election. The campaign targeted opposition candidates with false narratives to suppress voter turnout and influence public opinion. The incident highlights ongoing challenges for social media platforms in moderating government-linked disinformation campaigns in developing countries.
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.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to conduct large-scale disinformation campaigns, malicious surveillance, or targeted and sophisticated automated censorship and propaganda, with the aim of manipulating political processes, public opinion, and behavior.
Also classified as
The coordinated campaign created a fake news feedback loop that flooded the political discourse and sowed distrust among voters.
Social media platforms failed to adequately moderate and prevent repeat political disinformation campaigns in developing countries like Honduras.
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.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The disinformation campaign was intentionally designed and deployed by human political actors post-deployment to manipulate the Honduran election.
Added categories peak at 3: Substantial (Epistemic Harm, Inferred)
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.
The automated accounts performed a specific, pre-programmed function of posting identical tweets at the exact same second, operating as a simple automated tool rather than exhibiting autonomous planning or decision-making.
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.
The automated accounts were used to generate and amplify social media posts containing political opinions and conspiracy theories to influence public discourse.
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.
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.
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.
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.
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.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
