GRF · Information Integrity and Public Trust
AI generated information
Generative AI produces text, images, audio and video that can be hard to tell from human work. It lowers the cost of producing persuasive and false content, and of impersonating real people. Governance responds with provenance standards, labelling duties and platform rules, each with limits. This page sets out how AI-generated information affects integrity, the records that show it, and who decides.
How the work runs.
Narrative monitor
AI generated information · Crisis information cell · Week 34
Narratives
| Tap water poisoned after flood | 81 | High | False | Correction out |
| Shelters charging fees | 64 | Medium | False | Drafting |
| Dam about to fail | 69 | High | Checking | Holding line |
| Aid only for registered voters | 22 | Medium | False | Prebunked |
Spread
Response protocol
- Detectmonitor
- Verifyprovenance
- Assess harmthreshold
- Respondcorrect · prebunk
- Reporttransparency
Fig.Narrative monitor: example names and illustrative values, not a record or a live system.
AI generated information: how the risk develops.
Synthetic media and authenticity. The dossier below sets out how the risk develops, what is measured, where the data come from, how it is modelled and who decides.
Production cost collapses
Generating plausible text, images and voices is cheap and fast. Volume overwhelms review, and low-quality synthetic content fills search results and feeds.
Impersonation becomes easy
Voice and video clones of real people enable fraud, harassment and political deception. Victims include public figures and private individuals.
Detection is unreliable
Detectors produce false positives and miss content that has been edited or compressed. Reliance on detection alone can wrongly accuse genuine content.
Doubt devalues the real
When anything could be fake, genuine evidence can be dismissed as synthetic. This liar's dividend weakens documentation of real events.
How AI generated information is governed: instruments, mandates and commitments.
The instruments, mandates, actors and accountability arrangements a policymaker, diplomat or development agency works with, citing the agreements that set them. Structures follow the frameworks named. Any state or progress drawn is illustrative, not a record.
Accountability cycle
How commitments on AI generated information are held to account, from commitment to correction.
- CommitGovernmentA commitment with a mandate and a legal basis.
- PlanLead agencyA national plan with owners and milestones.
- ResourceMinistry of financeA budget line or financing on record.
- ImplementAgencies and local governmentDelivery against the plan.
- ReportStatistics officeIndicators reported in agreed definitions.
- ReviewLegislature, audit, civil societyIndependent review of results and gaps.
- CorrectGovernmentChanges to the plan, published with reasons.
What is measured, and how well.
Each indicator carries its unit, its cadence, the kind of source behind it and the evidence label a value would carry. No values are shown: a programme supplies them under its own authority.
| Indicator | Unit | Cadence | Source type | Label |
|---|---|---|---|---|
| Platforms implementing content credentials | count | on change | C2PA and platform records | DECLARED |
| Synthetic content labelling rules in force | count of jurisdictions | on change | Legislation | Published statistics |
| Reported deepfake incidents | count by harm type | monthly | Incident databases | REPORTED |
| Detection tool accuracy | true and false positive rates | per evaluation | Independent benchmarks | ESTIMATED |
| Election-related synthetic media incidents | count | per election | Monitoring reports | REPORTED |
| Public ability to identify synthetic media | share correct in tests | per study | Experimental studies | ESTIMATED |
Where the evidence comes from.
Named public datasets and institutional records, each with its licence class. Sources by agreement are used only under their terms and never republished.
| Source | What it gives | Licence class |
|---|---|---|
| C2PA specifications and conformance records | Provenance standard and adopters | Free, C2PA terms |
| OECD AI Incidents Monitor | Incidents including synthetic media harms | Free, OECD terms |
| AI Incident Database | Reported incidents involving AI | Open, licence to check |
| Partnership on AI Responsible Practices for Synthetic Media | Framework and case studies for synthetic media | Free to read, terms to check |
| Electoral authority and observer reports | Synthetic media in election campaigns | Public records, terms per publisher |
Who decides, and what the evidence serves.
Information Integrity and Public Trust prepares evidence. The decisions below are taken by the authorities named, under their own rules, and never by the platform.
| Who decides | The question | Evidence this page serves |
|---|---|---|
| Legislatures and AI regulators | What labelling and disclosure duties apply to synthetic content | Comparative rules and evidence on effects |
| Platforms and AI developers | Whether to attach provenance and how to label content | Standards and implementation practice |
| Electoral authorities | How to address synthetic media in campaigns | Monitoring reports and comparable rules |
| Newsrooms | How to verify media and disclose AI use | Verification practice and standards |
Outputs this topic produces
- Guide to content provenance, watermarking and labelling
- Comparative notes on synthetic media rules
- Dialogue briefs on the limits of detection
- Questions for electoral authorities on synthetic media
What experts ask.
Researchers on AI-generated information ask how to keep authentic evidence trustworthy.
- How is generative AI changing the volume, quality and cost of misleading content?
- How reliable are tools for detecting synthetic media, and how quickly do they become outdated?
- Can provenance and watermarking establish authenticity at scale?
- How does the possibility of synthetic media affect trust in genuine evidence?
- What obligations should apply to developers, deployers and distributors of generative AI?
Cheaper fakes and doubted truths.
Generative AI lowers the cost of producing convincing false content, including impersonation of public figures and fabricated evidence. It also enables personalised and multilingual content at scale. At the same time, the mere possibility of synthetic media allows genuine evidence to be dismissed as fake, a dynamic sometimes called the liar's dividend.
Detection tools can identify some synthetic content, but they are imperfect and engage in a continuing race with generation methods. Provenance approaches, which attach verifiable information about origin and editing to content, offer a complementary path, but depend on wide adoption and can be stripped. No single technical measure solves the problem; institutional practices and media literacy matter alongside technology.
Why now: generative AI has made realistic false text, images and audio cheap to produce, and labelling and provenance rules are still being adopted. The platform is designed to compare provenance and disclosure practice and to propose common practice to the Nexus Standards where institutions lack it.
How the evidence is built, and where it stops.
- 01
Detection research
Developing and evaluating classifiers with artificial intelligence and machine learning. Accuracy degrades as generators improve.
- 02
Provenance standards
Evaluating content credentials and watermarking. Adoption is incomplete and marks can be removed.
- 03
Behavioural research
Studying how people judge authenticity with social and behavioural science. People overestimate their ability to spot fakes.
- 04
Incident analysis
Studying real cases of synthetic media harm. Cases are often reported selectively.
How it is studied, and what the methods cannot say.
- 01Artificial intelligence and machine learningClassifier error in moderation · Synthetic media detection limits · Recommender auditsApplied here
- 02Social and behavioural scienceMeasuring trust and exposure · Testing interventions · Qualitative study of information needsApplied here
- 03Network science and contagionDetecting coordinated behaviour · Tracing narrative spread · Mapping media ecosystemsApplied here
Where this page leads across the Consortium.
Questions and answers.
What are content credentials?
Can AI-generated content be reliably detected?
What is the liar's dividend?
Can AI-generated content always be detected?
Is all AI-generated content harmful?
Contribute evidence from this page.
Reports is the Consortium's quarterly publication. What you know about this subject can become evidence, a case or a correction in the next edition, credited to you and published under CC BY 4.0 where it is your own work.
- EmailGateway GW-GRFSend it from this pageEvidence, a correction, a case, a method, a review, a translation or profile input. It arrives tagged with this page, so the editor who owns it receives it.
- ZenodoOpen depositDeposit it and submit to the communityPublish your own record on Zenodo with your ORCID iD, then submit it to the Reports community there. Curators review it; the DOI stays yours.
- ReportsGlobal Risks ForumSee every way to contributeThe contribution types, how each is reviewed, what credit you receive, the licence and what is never accepted.
Never send personal data, sensitive infrastructure detail or anything you have no right to share. A contribution is reviewed before use; sending it confers no authorship or editorial rights, and sponsorship never buys a finding.
Bring a governance question to the Forum.
Information and enquiries are open today. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.
Legal documents.
Show all 18 legal documents
Maps, data and figures on this site are shown for context and general information. They are not advice, forecasts or warnings; the limits of the information are set out in the Disclaimers and Data Notices.