Category 10 of 12 · Tool level
Synthetic or manipulated audio, video, imagery, or multimodal media is used to impersonate trusted people, fabricate evidence, provoke action, or undermine trust. Evidence trail for AIP-10-0376: 8 selected sources, 3 with a bounded automated source check.
Definition and boundary
Intentional use of synthetic or manipulated media to create false evidence, impersonate a person, provoke action, discredit authentic evidence, or degrade trust for a strategic objective. Evidence trail for AIP-10-0377: 8 selected sources, 3 with a bounded automated source check.
Disclosed satire, ordinary editing, accessibility dubbing, or synthetic media that is not used deceptively. Evidence trail for AIP-10-0378: 8 selected sources, 3 with a bounded automated source check.
Strategic and human significance
Audio and video can trigger immediate decisions before verification catches up. Evidence trail for AIP-10-0380: 8 selected sources, 3 with a bounded automated source check.
The mere possibility of deepfakes also enables the liar’s dividend, allowing authentic evidence to be dismissed as fake. Evidence trail for AIP-10-0381: 8 selected sources, 3 with a bounded automated source check.
Timing, trusted identity, and confirmation bias can matter more than perfect technical realism, especially in fast-moving crises. Evidence trail for AIP-10-0379: 8 selected sources, 3 with a bounded automated source check.
Change from pre-AI practice
Evidence maturity
The report documents election, conflict, school, and corporate cases involving confirmed synthetic audio or video, while emphasizing that effects are often difficult to isolate.
Public cases document synthetic media used in elections, conflict, fraud, and reputational attacks. Evidence trail for AIP-10-0387: 8 selected sources, 3 with a bounded automated source check.
High-quality synthetic media can be difficult for people to distinguish from authentic media. Evidence trail for AIP-10-0388: 8 selected sources, 3 with a bounded automated source check.
Reach and temporal association do not prove that a deepfake changed an election, military decision, or market outcome. Evidence trail for AIP-10-0389: 8 selected sources, 3 with a bounded automated source check.
Conflicting findings and scope boundaries
These are not errors to hide. They reflect different study designs, measures, time horizons, and operational contexts.
TENSION-PROVENANCE-01
Evidence position A: Content Credentials can provide cryptographically verifiable, tamper-evident provenance indicators.
Evidence position B: Credentials do not establish that depicted events are true, provenance can be incomplete, and metadata can be removed or separated from an asset.
Publication rule: Present provenance as one defensive signal alongside source verification, fact-checking, and forensics. Absence of a credential is not proof of fakery.
TENSION-LAW-01
Evidence position A: The reports identify existing enforcement actions and regulatory controls relevant to synthetic media, sensitive data, manipulation, and automated decision systems.
Evidence position B: Case posture, effective dates, definitions, constitutional constraints, and remedies vary across jurisdictions and change over time.
Publication rule: Date-stamp legal statements, identify the jurisdiction and procedural posture, and require current primary-source recheck before publication acceptance.
TENSION-LEGAL-CURRENTNESS-01
Evidence position A: The reports identify statutes, regulations, enforcement actions, and litigation relevant to manipulative or synthetic AI systems.
Evidence position B: Effective dates, amendments, injunctions, appeals, definitions, and remedies vary by jurisdiction and can change after the report date.
Publication rule: Require source-specific currentness review by an authorized human before publication acceptance.
Conceptual mechanisms
Voice cloning and fabricated communications. Evidence trail for AIP-10-0390: 8 selected sources, 3 with a bounded automated source check.
Face replacement and reenactment. Evidence trail for AIP-10-0391: 8 selected sources, 3 with a bounded automated source check.
Synthetic event footage. Evidence trail for AIP-10-0392: 8 selected sources, 3 with a bounded automated source check.
Hybrid fakes combining authentic and synthetic elements. Evidence trail for AIP-10-0393: 8 selected sources, 3 with a bounded automated source check.
False claims that authentic evidence is synthetic. Evidence trail for AIP-10-0394: 8 selected sources, 3 with a bounded automated source check.
These descriptions explain capability and risk. They intentionally omit procedures, targeting criteria, scripts, and evasion methods.
Evidence and examples
Evidence trail for AIP-10-0395: 1 selected sources, 0 with a bounded automated source check.
Evidence trail for AIP-10-0396: 1 selected sources, 1 with a bounded automated source check.
Case studies show documented events or bounded experiments. They do not establish prevalence, general causation, or guaranteed persuasive effect.
Risk and failure analysis
Detection and defense
Caveat: Genuine eyewitness material may also be unique. Evidence trail for AIP-10-0405: 8 selected sources, 3 with a bounded automated source check.
Caveat: Timing is contextual evidence, not proof of synthesis. Evidence trail for AIP-10-0406: 8 selected sources, 3 with a bounded automated source check.
Caveat: Out-of-band verification is stronger than perceptual artifacts alone. Evidence trail for AIP-10-0407: 8 selected sources, 3 with a bounded automated source check.
Governance and safeguards
Open questions
Sources and evidence boundary
Primary synthesis: Deepfake Psychological Operations Research. The complete report is retained in a non-public provenance directory with SHA-256 8dec5d4accf0bd6fea3d357e9648ae18d77b3939a7c628c1caf109ed2e47e09a.
Thirty high-impact references received bounded automated retrieval, official corroboration, or stronger-source substitution. All 91 selected references are used by the 501-claim citation graph, but the full report corpus and human editorial acceptance remain unverified. Source type labels are editorial classifications, not quality scores.
Metadata is resolved, but the linked source has not received this release's independent content-scope check.
The 2026 preprint presents an independent critical security analysis of C2PA. Current C2PA documentation itself states that credentials are not truth judgments, may be incomplete, and can be removed from an asset.
Limits: The critical paper is a preprint and should not be treated as settled consensus. The official standard and the critical analysis address different questions and should be presented together.
The paper evaluates robustness limits and enhancement methods for SynthID-style text watermarking, supporting the claim that watermark performance depends on transformations and evaluation conditions.
Limits: The source is a preprint and addresses one watermarking approach rather than all provenance systems. Watermark removal or degradation does not prove malicious intent.
Metadata is resolved, but the linked source has not received this release's independent content-scope check.
Metadata is resolved, but the linked source has not received this release's independent content-scope check.
Metadata is resolved, but the linked source has not received this release's independent content-scope check.
Metadata is resolved, but the linked source has not received this release's independent content-scope check.
Baltimore County Police reported the arrest and charges associated with the Pikesville High School AI-generated audio case.
Limits: A charging announcement is not a final adjudication of guilt.