Network of 171 AI-Powered Bots Reportedly Spread Political Disinformation Ahead of Ghana’s December 2024 General Election

A network of 171 bot accounts on X are alleged to have used ChatGPT to generate political content supporting Ghana’s New Patriotic Party (NPP) and its presidential candidate, Mahamudu Bawumia, ahead of the December 2024 election. The AI-generated posts reportedly praised Bawumia while spreading disinformation targeting the opposition candidate, John Mahama, of the National Democratic Congress (NDC).

Risk classification

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

The incident involved a coordinated network of AI-powered bots used to conduct a political influence campaign to manipulate public opinion and democratic processes in Ghana.

Additional risk subdomains

• 3.1 False or misleading information: The bot network generated false and disparaging claims accusing the opposition candidate of being a drunkard.
• 6.5 Governance failure: The campaign was enabled by X's failure to maintain adequate content moderation and trust and safety oversight.

Causal factors

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

The risk was caused by human bad actors who intentionally deployed a network of ChatGPT-powered bots to influence the Ghanaian election after the AI model was deployed.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Risk Level 3: Limited Risk. The system involved is ChatGPT, a chatbot used to generate text content, which falls under the category of AI systems that require transparency obligations so users are aware they are interacting with AI-generated content.

AI system and alleged parties

• AI system: ChatGPT
• AI purpose: Social Media Content Generation; Writing Assistant
• Behaviour type: Assistant
• Alleged developer: OpenAI
• Alleged deployer: Pro-New Patriotic Party bot network
• Alleged harmed parties: National Democratic Congress, John Mahama, Ghanaian 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 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 Minor
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Malicious content

Reported: The report explicitly describes the creation and spread of partisan political content and disparaging posts targeting an opposition candidate.

Directly caused: A network of 171 bot accounts used ChatGPT to generate posts promoting the NPP and disparaging opposition candidate John Mahama, accusing him of being a drunkard.

Inferred additional harm: It is likely that similar AI-generated political bot networks are operating undetected on X and other platforms to spread malicious or toxic political content.

Democracy

Reported: The report explicitly describes the use of AI to secretly influence democratic elections in Ghana.

Directly caused: The bot network used ChatGPT to generate and amplify partisan political messaging to influence the December 7 presidential election.

Indirectly caused: The proliferation of AI-generated bot networks on X undermines the integrity of public political discourse and democratic processes.

Inferred additional harm: The covert use of AI to manipulate elections could lead to a broader erosion of trust in democratic institutions and election outcomes in Ghana and globally.

Epistemic

Reported: The report explicitly describes the spread of unsubstantiated claims and partisan propaganda by AI-generated bot accounts.

Directly caused: The 171 bot accounts posted AI-generated content accusing opposition candidate John Mahama of being a drunkard and promoting NPP talking points.

Indirectly caused: The reduction in X's content moderation allowed verified accounts and bots to spread misinformation with greater prominence, misleading users.

Inferred additional harm: The widespread use of AI to generate realistic political content at scale makes it increasingly difficult for voters to distinguish between genuine public opinion and automated propaganda, degrading the shared information ecosystem.

People affected

• Occurrences reported: 1
• People reportedly harmed: 1
• People reportedly exposed: 100000

Potential causes

Management

• Prioritization of Cost Cuts: X management cut 80% of safety engineers, reducing moderation capacity.
• Monetization of Verification: Selling blue checks allowed bad actors to boost bot visibility.
• Inadequate Abuse Prevention: OpenAI and X failed to proactively prevent political influence campaigns.

Technology

• ChatGPT Text Generation: Allowed easy, automated creation of realistic political posts at scale.
• AI-Generated Profile Photos: Created fake human personas to make bot accounts look authentic.
• Algorithmic Amplification on X: X's algorithms prioritized verified bot content for higher visibility.

Data Inputs

• Existing Political Narratives: Input of existing party talking points allowed bots to mimic real debates.
• Lack of Content Labeling on X: Absence of labels identifying AI-generated content or automated accounts.

Human Factors

• Public Trust in Blue Checks: Users believed blue checkmarks indicated reliable and verified accounts.
• Bad Actors' Intent to Influence: Malicious actors deliberately leveraged AI tools to manipulate elections.
• User Susceptibility: Users struggled to distinguish AI-generated bot accounts from real people.

Process and Methods

• Diminished Content Moderation: X reduced trust, safety, and election integrity teams, weakening defense.
• Predictable Posting Patterns: Bots posted at highly regular intervals, revealing automated behavior.
• Ineffective Reporting Mechanisms: Slow response from platforms to reports of coordinated bot activity.

Regulatory Environment

• Lack of AI Content Regulations: No legal frameworks governing AI-generated content in political campaigns.
• Weak Platform Oversight: Regulatory bodies lacked authority to enforce election integrity on X.

Information quality

• Classification confidence: High
• Reason for confidence: The report is highly detailed, citing a specific investigation by NewsGuard that identified 171 bot accounts on X using ChatGPT. The role of the AI system (ChatGPT) is explicitly stated, and the nature of the campaign (pro-NPP, anti-NDC) is well-documented with examples of posts and hashtags. There is little ambiguity about the events, though the exact real-world impact on voters is noted as difficult to measure.
• Ambiguities identified: The exact real-world impact of the bot network on Ghanaian voters' decisions is difficult to disentangle from organic social media activity.

A network of 171 bot accounts on X used ChatGPT to generate and spread pro-government propaganda and defame an opposition candidate ahead of Ghana's presidential election. While highlighting the growing ease of deploying AI for localized political influence, the incident represents a minor national security threat with no compromise of physical systems, critical infrastructure, or core government operations.

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

Threat characteristics

• Imminence: Near-term. The campaign was active leading up to the national election, representing a developing situation that required active monitoring and response.
• Autonomy: Human-supervised. The bot network used ChatGPT to autonomously generate and post content at scale, but operated under the strategic direction and setup of human actors.
• Novelty: Evolved capability. Represents an evolution of traditional bot-driven disinformation campaigns by integrating generative AI to produce unique, persuasive political content at scale.

Impact by dimension

• Physical security: Negligible. No physical systems, critical infrastructure, or human safety threats were involved or impacted in this incident.
• Information security: Minor. Coordinated political disinformation campaign using 171 ChatGPT-powered bots to influence Ghana's election. However, the scale was relatively small and did not compromise state intelligence or classified systems.
• Sovereignty: Minor. The campaign targeted the presidential election, attempting to influence voters. However, there was no compromise of actual electoral systems, voting infrastructure, or core government decision-making processes.
• Economic security: Negligible. No threats to financial systems, strategic technologies, or economic stability were reported or indicated in this incident.
• Societal stability: Minor. The bot network spread defamatory content against an opposition candidate, potentially polarising the electorate, but did not result in mass civil unrest, systemic discrimination, or widespread human rights violations.