AI-Enabled Deepfake Psychological Operations

A · DEFINITION

What this category means

Intentional use of synthetic or manipulated audio, video, imagery, or multimodal media to alter perceptions, impersonate trusted figures, fabricate evidence, provoke action, discredit authentic evidence, or undermine trust.

Outside its scope

Deepfakes are a subset of synthetic media. Mislabelled authentic media, cheapfakes, and hybrid edits may produce similar psychological effects and must be distinguished during verification.

Evidence basis^{[1, 2]}

B · WHY IT MATTERS

Strategic and public-interest significance

Audio and video are often treated as direct evidence. Synthetic media can exploit that trust during elections, armed conflict, financial transactions, or communal tension, and can also allow wrongdoers to deny genuine recordings.

Primary AI role: Tool
Unit of influence: individual / group / institution
Degree of autonomy: Low; distribution may be automated
Evidence maturity: Documented current use

Evidence basis^{[1, 2, 6]}

C · HOW AI CHANGES IT

What changes compared with pre-AI practice

Voice cloning, facial reenactment, and generative video reduce the resources required to impersonate people or fabricate events. Post-hoc detection remains an arms race, so source authentication and out-of-band verification are often more reliable than visual inspection.

Evidence basis^{[1, 3, 4]}

D · CAPABILITY STATUS

Separate current evidence from prospective risk

Confirmed real-world use

Documented incidents include political audio, wartime surrender video, corporate fraud, school defamation, and voter-suppression robocalls.

Demonstrated technical capability

Humans often perform near chance when judging high-quality synthetic media; forensic systems also fail under compression and novel generators.

Plausible near-term development

Real-time interactive impersonation in calls and meetings is an expanding risk.

Speculative or unsupported claims

No detector or provenance standard currently guarantees universal authenticity.

Evidence basis^{[1, 2, 3, 4, 5, 6]}

E · KEY MECHANISMS

Conceptual mechanisms—not procedures

01

Voice cloning and fabricated communications.

02

Face replacement, reenactment, or synthetic events.

03

Hybrid edits embedded in mostly authentic material.

04

False claims that genuine media is AI-generated.

Evidence basis^{[1, 2]}

F · EVIDENCE AND EXAMPLES

What occurred, what is known, and what remains unknown

Reach, engagement, and visibility are not treated as proof of persuasion or behavior change.

Compare every qualified case across the taxonomy

Zelensky surrender video^{[1, 3]}

What occurred: A synthetic video depicted Ukraine’s president instructing troops to surrender after a news channel was compromised in March 2022.
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.

Pikesville High School principal audio^{[1, 4]}

What occurred: A school employee allegedly created a cloned recording that falsely portrayed the principal making racist and antisemitic remarks.
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.

Hong Kong video-conference fraud^[1]

What occurred: A finance employee transferred about $25 million after a meeting featuring apparent deepfake colleagues.
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.

Ali Bongo video and the liar’s dividend^{[1, 5]}

What occurred: An authentic video of Gabon’s president after a stroke was widely alleged to be synthetic.
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.

G · RISKS AND FAILURE MODES

Malicious-use risks and reasons the capability may fail

Primary risks

False media can trigger panic, fraud, retaliation, or reputational harm before correction.
False positives can discredit authentic evidence and harm innocent creators.
Crisis timing can make even low-quality media effective.

Evidence basis^{[1, 2, 5, 4]}

Limits and failure modes

Longer synthetic video may reveal temporal and physical inconsistencies.
Detection accuracy degrades under recompression, cropping, noise, and unseen generators.

Evidence basis^{[1, 3]}

H · DETECTION AND DEFENSIVE INDICATORS

Signals are suggestive, not automatic proof

False-positive warning: No single detector score, writing style, profile image artifact, posting pattern, or political similarity should be used alone to accuse a person or organization.

Inconsistencies in lighting, acoustic environment, or motion can support review but are not proof.
The highest-value check is often whether the claimed speaker or institution can authenticate the communication through another channel.

Evidence basis^{[1, 3, 4]}

I · GOVERNANCE AND SAFEGUARDS

Layered controls, oversight, and accountability

Establish prearranged out-of-band verification for financial and emergency instructions.
Use a truth-first crisis response that avoids repeatedly amplifying the false content.
Combine provenance, forensic review, contextual evidence, and chain of custody.

Evidence basis^{[1, 6, 2]}

Related cross-category safeguards

The resilience guide compares these controls with their limits and evidence context across the full taxonomy.

Media literacy and pre-bunking Source verification and crisis procedures Provenance and authenticity signals Careful detection, due process, and false-positive controls Institutional communication and trust repair

Legal conclusions depend on jurisdiction and facts; this page summarizes the corresponding report and is not legal advice.

J · RESEARCH GAPS

Questions the evidence does not yet resolve

Effectiveness of labels and prebunking across cultures.
Privacy-preserving provenance that survives distribution.
Legal standards for authenticating synthetic or disputed evidence.

Evidence basis^{[1, 5]}

Compare this category’s questions across the research agenda

K · SOURCES

Traceable source list

The commissioned report is the organizing source. External records below are the principal sources retained for the public synthesis; source quality varies by type and is labelled.

English and Spanish editions are published from the same structured record. Bilingual parity is validated for every release; source titles may remain in their original publication language.

Browse the cross-category case register Browse the cross-category evidence register Compare the terminology used across categories Compare cross-category resilience safeguards Review the cross-category open research agenda

Commissioned research report AI-Enabled Deepfake Psychological Operations
Evidence links: 14
Legal scholarship Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security
Evidence links: 6
Incident database record AI Incident 573: Zelensky surrender deepfake
Evidence links: 5
Secondary reporting School principal was framed using AI-generated racist rant
Evidence links: 5
Peer-reviewed research Deepfakes and disinformation: Exploring the impact of synthetic political video
Evidence links: 4
Peer-reviewed legal research Emerging need to regulate deepfakes in international law
Evidence links: 3