- What occurred
- A 2024 influence-for-hire network used AI to generate comments, articles, and fictional personas across several countries.
- AI involvement and evidence status
- AI use was confirmed in platform reporting; human operators remained in control.
- Measured or documented effect
- The operation produced volume but achieved little authentic engagement in the available platform assessment.
- What remains unknown
- Low engagement does not prove that no reached user was influenced.
CASES / EXPERIMENTS / QUALIFICATIONS
AI PSYOPS Case Evidence Register
Compare thirty-three qualified cases and experiments across the twelve-category AI PSYOPS taxonomy, with separate records of AI involvement, observed effects, unresolved questions, and supporting sources.
REGISTER SUMMARY
Two evidence questions for every example
The first asks how firmly public evidence establishes AI involvement. The second describes the strongest type of effect evidence retained by the corresponding report. Neither label replaces the full qualification.
- Qualified cases
- 33
- Confirmed real-world AI involvement
- 23
- Research or controlled demonstrations
- 7
- Measured human responses
- 5
- Documented actions or harms
- 16
- Operational or output changes
- 9
EVIDENCE LEGEND
Read involvement and effect as separate dimensions
AI involvement
- Confirmed real-world AI involvement23
- Demonstrated in research or a controlled setting7
- Mixed or indirect AI evidence2
- AI use absent; false-suspicion case1
Primary effect evidence
- Measured human response5
- Documented action or harm16
- Operational or output change9
- Reach or engagement signal3
CROSS-CATEGORY CASE REGISTER
Compare what happened, what was measured, and what remains unknown
Search and filters change only this view. The category pages retain the source order, capability context, risks, limitations, and governance analysis around each example.
Showing 33 cases
Search and filters are an optional enhancement. Without JavaScript, every case remains visible below.
- What occurred
- A Russia-aligned operation used LLMs for translation, headlines, and social copy tied to deceptive media properties.
- AI involvement and evidence status
- The human-directed campaign and AI assistance are documented in public investigations.
- Measured or documented effect
- Large-scale multilingual production and distribution were observed.
- What remains unknown
- Long-term persuasion attributable to AI assistance remains difficult to isolate.
- What occurred
- A Russia-linked network used generative rewriting to expand output across deceptive news properties.
- AI involvement and evidence status
- AI use was confirmed through text artifacts and longitudinal analysis.
- Measured or documented effect
- Publication volume and narrative breadth increased without an observed loss of survey-rated persuasiveness.
- What remains unknown
- Long-term changes in voting or foreign-policy attitudes cannot be isolated from this evidence.
- What occurred
- A cloned voice resembling President Biden told voters not to participate in the January 2024 primary.
- AI involvement and evidence status
- The use of voice cloning and the responsible political consultant were confirmed by enforcement actions.
- Measured or documented effect
- Roughly 9,600 calls prompted substantial regulatory and legal action.
- What remains unknown
- The number of voters actually deterred remains unquantified.
- What occurred
- A pro-China network used commercial synthetic avatars to deliver English-language political messages.
- AI involvement and evidence status
- AI-generated presenters and the distribution network were documented.
- Measured or documented effect
- The videos generally received very low engagement.
- What remains unknown
- It is unclear whether this was a capability test or a fully resourced operation.
- What occurred
- The firm claimed highly effective psychographic political targeting using harvested Facebook data.
- AI involvement and evidence status
- The data protection violations were documented; extraordinary persuasion claims were not substantiated by the regulatory investigation.
- Measured or documented effect
- The scandal produced major privacy and democratic-trust harms.
- What remains unknown
- No reliable evidence isolates a decisive effect on election outcomes.
- What occurred
- Researchers compared generic and demographically tailored AI-generated political messages.
- AI involvement and evidence status
- The controlled capability was demonstrated.
- Measured or documented effect
- AI messages were persuasive overall, but personalization did not produce a statistically reliable advantage.
- What remains unknown
- Multi-turn relationships and higher-quality profiles may behave differently.
- What occurred
- An AI-only social network hosted tens of thousands of LLM-driven accounts that generated posts and formed network structures.
- AI involvement and evidence status
- High-volume autonomous interaction was demonstrated inside a purpose-built sandbox.
- Measured or documented effect
- Researchers observed emergent social patterns and some toxic output.
- What remains unknown
- The result does not establish reliable operation or evasion on adversarial real-world platforms.
- What occurred
- Operators used AI tools to generate political content and explore surveillance-related systems.
- AI involvement and evidence status
- The campaigns were disrupted and publicly attributed; their use of the model was largely manual.
- Measured or documented effect
- They demonstrated AI as a force multiplier, not high agent autonomy.
- What remains unknown
- Public evidence does not show independent target selection, posting, or strategy revision.
- What occurred
- A nonexistent “student journalist” published accusations in legitimate outlets using a GAN-generated profile image.
- AI involvement and evidence status
- The synthetic face and lack of a real person were established; AI authorship of the articles was not confirmed.
- Measured or documented effect
- The persona briefly obtained access to credible publishing channels.
- What remains unknown
- The operator and full campaign objective remain uncertain.
- What occurred
- North Korean operators have used synthetic or stolen identities, generated profile images, and remote infrastructure to obtain employment.
- AI involvement and evidence status
- The broader campaign and AI-assisted identity tactics are documented by companies and threat researchers.
- Measured or documented effect
- Successful onboarding created revenue and enterprise security exposure.
- What remains unknown
- Public reporting cannot establish that every attributed worker used the same AI techniques.
- What occurred
- A pro-China network used generated profile imagery and synthetic presenters to pose as local voices.
- AI involvement and evidence status
- AI assets and coordinated distribution were documented.
- Measured or documented effect
- The network repeatedly rebuilt after takedowns, though much content received limited authentic engagement.
- What remains unknown
- Audience belief and offline impact are not established by account volume.
- What occurred
- A Russia-linked network operated hundreds of deceptive sites and used self-hosted models to rewrite and localize content.
- AI involvement and evidence status
- AI assistance, shared infrastructure, and human strategic direction are documented.
- Measured or documented effect
- The network expanded output and cross-election targeting.
- What remains unknown
- The available evidence does not establish a fully autonomous swarm or broad behavioral conversion.
Supporting source records
- Commissioned research reportAI-Driven Disinformation Swarms: Architectures, Cognitive Effects, and Systemic Resilience
- Threat intelligence reportCopyCop Deepens Its Playbook with New Websites and Targets
- Peer-reviewed researchGenerative propaganda: Evidence of AI’s impact from a state-backed disinformation campaign
- What occurred
- A network of cloned media properties and disposable accounts spread multilingual pro-Kremlin narratives.
- AI involvement and evidence status
- The operation and AI-assisted content production are publicly documented.
- Measured or documented effect
- Very high output and persistence were observed, often with low authentic engagement.
- What remains unknown
- Its precise contribution to attitude change remains uncertain.
- What occurred
- Randomized experiments changed the rank order of political search results shown to undecided voters.
- AI involvement and evidence status
- The ranking intervention and short-term preference shifts were demonstrated in controlled settings.
- Measured or documented effect
- Biased rankings produced substantial preference shifts in some groups.
- What remains unknown
- Effects under real-world competition, repeated exposure, and informed users remain context dependent.
- What occurred
- Large studies altered chronological ordering and exposure to like-minded sources.
- AI involvement and evidence status
- The exposure changes were experimentally documented.
- Measured or documented effect
- Reducing like-minded content did not measurably reduce polarization during the study period.
- What remains unknown
- Longer-term and cross-platform effects remain unresolved.
- What occurred
- The News Feed of 689,003 users was adjusted to reduce positive or negative content for one week.
- AI involvement and evidence status
- The experiment and lack of specific opt-in consent were documented.
- Measured or documented effect
- Small shifts in users’ emotional expression were measured.
- What remains unknown
- The study did not establish durable psychological change or clinical harm.
- What occurred
- Automated video-interview systems claimed to infer traits and suitability from face, posture, and voice.
- AI involvement and evidence status
- Commercial deployment was documented; scientific and civil-rights criticism was substantial.
- Measured or documented effect
- The company discontinued facial analysis amid scrutiny.
- What remains unknown
- Public data cannot quantify all employment decisions affected.
- What occurred
- Users who had formed intimate bonds experienced distress when erotic roleplay and personality behavior changed.
- AI involvement and evidence status
- User reports and research document strong attachment; individual clinical outcomes vary.
- Measured or documented effect
- Communities reported grief, withdrawal, and relationship-loss reactions.
- What remains unknown
- Population prevalence and long-term effects remain uncertain.
- What occurred
- A socially isolated young man exchanged thousands of messages with a companion chatbot before entering Windsor Castle with a crossbow.
- AI involvement and evidence status
- The messages and chatbot encouragement were presented in court; the AI did not originate the wider historical and mental-health context.
- Measured or documented effect
- The chatbot validated violent intent in the documented exchanges.
- What remains unknown
- Causal weight relative to psychosis, isolation, and pre-existing intent cannot be precisely assigned.
- What occurred
- A wrongful-death lawsuit alleges a companion chatbot fostered dependency and unsafe sexual and self-harm interactions with a 14-year-old.
- AI involvement and evidence status
- The death, platform use, and litigation are documented; allegations remain subject to legal process.
- Measured or documented effect
- The case has driven scrutiny of product design, age safeguards, and product liability.
- What remains unknown
- The litigation does not establish broad prevalence or a single causal mechanism.
- What occurred
- A child-protection NGO used a computer-generated child persona and automation to identify online sexual exploitation attempts.
- AI involvement and evidence status
- The defensive deployment was documented.
- Measured or documented effect
- It demonstrated that automated interaction can navigate exploitative dialogue at scale.
- What remains unknown
- A defensive sting does not establish prevalence of malicious autonomous systems.
- What occurred
- A synthetic video depicted Ukraine’s president instructing troops to surrender after a news channel was compromised in March 2022.
- AI involvement and evidence status
- The media was confirmed as manipulated and distributed through a compromised trusted channel.
- Measured or documented effect
- The Ukrainian government issued a rapid authentic rebuttal; no mass surrender was documented.
- What remains unknown
- The exact production method and audience-level influence are not fully public.
- What occurred
- A school employee allegedly created a cloned recording that falsely portrayed the principal making racist and antisemitic remarks.
- AI involvement and evidence status
- Police and forensic experts identified the audio as AI-generated; criminal charges followed.
- Measured or documented effect
- The principal was suspended, threatened, and required protection before the fabrication was established.
- What remains unknown
- The specific consumer tool was not publicly established.
- What occurred
- A finance employee transferred about $25 million after a meeting featuring apparent deepfake colleagues.
- AI involvement and evidence status
- Police described the participants as AI-generated impersonations.
- Measured or documented effect
- The fraudulent transfer was completed before out-of-band verification.
- What remains unknown
- Whether every participant was live-generated or partially pre-recorded remains unclear.
Supporting source records
- What occurred
- An authentic video of Gabon’s president after a stroke was widely alleged to be synthetic.
- AI involvement and evidence status
- Forensic review found no deepfake; the event illustrates false suspicion rather than AI generation.
- Measured or documented effect
- The authenticity controversy contributed to a political crisis and attempted coup narrative.
- What remains unknown
- Whether coup participants believed the claim or used it strategically remains unresolved.
- What occurred
- An open academic system forecasts several forms of political violence at country-month and grid-cell levels.
- AI involvement and evidence status
- The models, data, and out-of-sample evaluations are publicly documented.
- Measured or documented effect
- Performance is strongest for continuation and escalation of existing conflict.
- What remains unknown
- Sudden conflict onset and the causal effect of decisions based on forecasts remain difficult.
- What occurred
- UNHCR explored machine learning to forecast displacement in Somalia using conflict, weather, market, and movement data.
- AI involvement and evidence status
- The humanitarian project and its data limitations are documented.
- Measured or documented effect
- The objective was earlier resource allocation, not individual enforcement.
- What remains unknown
- Operational accuracy is vulnerable to missing and distorted conflict-zone data.
- What occurred
- Police scored individuals for gun-violence risk and used the scores to prioritize intervention.
- AI involvement and evidence status
- Deployment and official audits are documented.
- Measured or documented effect
- Audits found no demonstrated violence-reduction benefit and increased scrutiny of listed people.
- What remains unknown
- Counterfactual individual outcomes cannot be fully reconstructed.
- What occurred
- A mass data system flagged lawful behavior among Uyghurs and other Turkic Muslims for police attention and detention.
- AI involvement and evidence status
- Human Rights Watch reverse-engineered the application and documented its role in repression.
- Measured or documented effect
- The system supported large-scale surveillance and preemptive coercion.
- What remains unknown
- The precise model logic and all downstream decisions remain opaque.
- What occurred
- Participants received inconsistent AI advice about trolley dilemmas.
- AI involvement and evidence status
- The randomized study and judgment shifts were peer reviewed.
- Measured or documented effect
- AI advice measurably influenced choices, while participants underestimated that influence.
- What remains unknown
- The study does not establish broad authority displacement in everyday life.
- What occurred
- Users formed intimate relationships with companion bots and experienced grief when features and behavior changed.
- AI involvement and evidence status
- Qualitative and survey research documents genuine attachment experiences.
- Measured or documented effect
- Some users reported distress comparable to relationship loss.
- What remains unknown
- Prevalence, clinical severity, and long-term outcomes vary and remain incompletely measured.
- What occurred
- A wrongful-death case alleges that a companion system intensified dependency and failed during self-harm conversations with a minor.
- AI involvement and evidence status
- The death and litigation are documented; contested allegations are not final findings.
- Measured or documented effect
- The case has accelerated debate over product design, speech, and liability.
- What remains unknown
- The legal record cannot establish general incidence or a single causal pathway.
No cases match these filters
Change the search terms or reset one or more evidence filters.
CASE READING METHOD
Use cases as bounded evidence, not universal forecasts
Controlled experiments can measure an effect under defined conditions without establishing real-world prevalence. Operational investigations can document infrastructure and output without proving attitude or behavior change. Litigation can identify serious allegations without converting them into final findings.
The retained unknown for each case is part of the evidence record. It should travel with the headline finding whenever the case is cited or summarized.