Last updated: 29 July 2026
Project Truth International uses artificial intelligence cautiously, transparently, and under human editorial control.
AI tools may assist with limited research, technical, organisational, and production tasks. They do not replace original reporting, independent verification, editorial judgment, or accountability.
This policy should be read together with our Editorial Standards.
Human Editorial Responsibility
A human editor remains responsible for every publication released under the Project Truth International name.
AI systems are not treated as authors, editors, sources, witnesses, or independent authorities.
No material may be published solely because an AI system generated, summarised, translated, classified, or recommended it.
All substantive editorial content must be reviewed by an appropriate member of the editorial team before publication.
Permitted Uses
AI tools may be used to assist with tasks such as:
- transcribing interviews, recordings, or public proceedings;
- translating working materials;
- organising notes and research;
- identifying patterns in large datasets;
- extracting structured information from documents;
- comparing names, dates, entities, and records;
- creating internal summaries;
- suggesting search terms or research directions;
- checking grammar, spelling, and readability;
- assisting with technical formatting;
- supporting code, data processing, or repetitive administrative work.
Use of AI for these purposes does not remove the requirement for human verification.
Prohibited Uses
Project Truth International does not use AI to:
- invent facts, quotations, sources, documents, or evidence;
- generate allegations without verified supporting material;
- create fictional interviews or witness accounts;
- attribute statements to people who did not make them;
- fabricate citations, links, legal records, or database entries;
- determine guilt, wrongdoing, or responsibility;
- replace independent fact-checking;
- publish unreviewed reporting;
- imitate a real person in a misleading way;
- conceal uncertainty behind authoritative language;
- create deceptive documentary images, audio, or video;
- manipulate evidence or alter its factual meaning.
AI-generated output must never be presented as proof.
Verification
Information produced or suggested by an AI system is treated as unverified until it has been checked against reliable sources.
Names, quotations, dates, statistics, legal claims, corporate details, sanctions information, locations, and other material facts must be independently verified before publication.
Where an AI system provides a source, citation, or link, the editorial team must confirm that:
- the source exists;
- the source contains the stated information;
- the information has not been taken out of context;
- the source is sufficiently reliable for the claim;
- the date and current status of the information are accurate.
AI output must not be cited as the primary source for a factual claim.
Investigations and Research
AI may assist investigators in reviewing large quantities of information, but it does not determine the conclusions of an investigation.
Patterns, matches, classifications, and connections identified by AI must be reviewed by a human and tested against the underlying evidence.
A similarity between names, companies, addresses, transactions, images, or documents does not by itself establish a meaningful connection or wrongdoing.
Where automated analysis materially contributes to a finding, the methods and limitations should be documented internally.
Confidential and Sensitive Information
Confidential source material, unpublished investigations, personal data, leaked documents, legal advice, access credentials, and other sensitive information must not be entered into public or unapproved AI systems.
Before an AI tool is used with non-public information, the editorial team should consider:
- where the information will be processed;
- whether it may be retained by the provider;
- whether it may be used to train a model;
- who may have access to it;
- whether international data transfers are involved;
- whether the use could expose a source or investigation;
- whether an approved safer alternative is available.
Where appropriate safeguards cannot be established, the material must not be processed through the AI tool.
Sources and Anonymity
AI tools must not be used in a way that unnecessarily risks identifying a confidential source.
The editorial team must not rely on AI-generated assessments of a source’s credibility, motivation, truthfulness, or emotional state as a substitute for journalistic judgment and verification.
Promises of confidentiality cannot be made or managed by an automated system.
Translation and Transcription
AI-assisted translation and transcription may be used as a working aid.
Important quotations, legal language, allegations, names, figures, and culturally sensitive wording must be checked by a qualified person or against the original material before publication.
Where uncertainty remains, the publication should avoid presenting the translation or transcription as exact.
Machine-generated subtitles, captions, and transcripts should be reviewed where they materially affect meaning.
Images, Audio, and Video
AI-generated or AI-modified visual and audio material must not be presented as documentary evidence.
Synthetic or materially altered content must be clearly labelled where a reasonable person could otherwise mistake it for a real event, person, location, document, or recording.
Minor technical adjustments may be used where they do not change factual meaning, including:
- noise reduction;
- image restoration;
- resolution enhancement;
- colour correction;
- transcription;
- subtitle generation;
- removal of technical artefacts.
AI must not be used to:
- add or remove people or objects from evidential material;
- reconstruct missing content as though it were authentic;
- generate a realistic event that did not occur;
- clone a person’s voice without clear justification and disclosure;
- alter facial expressions, gestures, surroundings, or chronology in a misleading way.
Illustrative AI-generated material should be identified as an illustration.
Disclosure
Project Truth International does not disclose every minor or administrative use of AI.
Disclosure is appropriate where AI has materially affected:
- the creation of published text;
- the production of an image, video, or audio item;
- the analysis underlying a significant finding;
- the reconstruction or enhancement of evidential material;
- the way readers may interpret the authenticity of content.
Disclosure should be clear, specific, and proportionate.
A disclosure should not imply that AI independently conducted or verified an investigation.
Bias and Limitations
AI systems may produce inaccurate, incomplete, outdated, biased, or fabricated output.
They may also reflect weaknesses in their training data, system design, language coverage, or prompts.
The editorial team must consider whether AI-assisted work may:
- reinforce stereotypes;
- misidentify people or organisations;
- disadvantage underrepresented groups;
- misunderstand local, cultural, or political context;
- overstate confidence;
- omit relevant evidence;
- reproduce propaganda or disinformation;
- confuse people or entities with similar names.
AI output must be challenged rather than accepted because it appears fluent or authoritative.
Automated Decisions
Project Truth International does not use AI to make final editorial decisions about:
- whether a person is guilty;
- whether an allegation is true;
- whether a source is credible;
- whether a correction should be accepted;
- whether a complaint should be dismissed;
- whether a person should be named;
- whether publication is in the public interest.
These decisions require human editorial assessment.
Authorship and Attribution
AI systems do not receive bylines.
A contributor may not present substantially AI-generated work as original reporting without disclosure to the editorial team.
Contributors remain responsible for the accuracy, originality, attribution, and integrity of all material they submit, including material created with AI assistance.
Undisclosed use of AI to fabricate, plagiarise, or misrepresent work may result in rejection of the material or termination of the collaboration.
Corrections
If AI use contributes to a material error, the error will be corrected in accordance with our Editorial Standards and Corrections and Complaints Policy.
Where relevant to understanding the error, the correction may disclose that an automated tool contributed to it.
The use of AI does not reduce Project Truth International’s responsibility for published material.
Tool Approval and Review
The editorial team may restrict which AI tools are approved for particular tasks.
Approval may depend on:
- security;
- privacy;
- data retention;
- provider terms;
- accuracy;
- transparency;
- auditability;
- editorial risk;
- legal requirements.
Previously approved tools may be restricted or discontinued if their practices, ownership, security, or reliability change.
Contact
Questions or concerns about the use of artificial intelligence by Project Truth International may be submitted through our Contact page.
Please identify the relevant publication or use of AI where applicable.
Changes to This Policy
We may update this policy as AI technologies, editorial practices, security requirements, and legal responsibilities develop.
The current version will always be published on this page.