Investigative Journalism Standards

Purpose and Threshold

Investigative journalism serves the public by bringing to light information that is concealed, difficult to access, poorly examined or not adequately understood. It can uncover wrongdoing, institutional failures, abuse of authority, conflicts of interest, misuse of resources, public deception, systemic risks or conduct that may have significant consequences for individuals and communities.

DesiAI uses the label Investigation only when an article contains substantial original reporting or evidence independently developed and verified by its editorial team or contributors. An article does not qualify as an investigation simply because it is lengthy, critical, based on numerous public sources or written in a formal style. Investigative work requires a higher level of planning, evidence verification, fairness, legal awareness, source protection, harm assessment and post-publication responsibility than ordinary commentary or general analysis.

Scope

These standards apply to investigations involving governments, public institutions, companies, political organizations, technology platforms, employers, financial activity, public-interest litigation, safety failures, corruption, privacy, surveillance, discrimination, environmental harm, health claims, organized manipulation, media systems, public expenditure and individuals exercising significant influence or authority. The same standards apply whether the work is prepared by a staff author, guest contributor, external researcher, collaborative team or specialist.

An article may be classified as an investigation where it combines a substantial amount of original evidence, such as original documents, independently obtained records, interviews, direct observation, technical testing, data analysis, financial reconstruction, documented source testimony, comparison of conflicting evidence, original chronology, verification of previously unconfirmed claims or disclosure of a matter with demonstrable public value.

The Investigation label is not used for articles that primarily summarize another publication, compile public statements without independent verification, reproduce material supplied by an organization, repeat litigation allegations, rely on a single interested source without meaningful corroboration, turn criticism of a company into an investigative presentation, present opinion as discovery or treat search results as the principal evidence.

Investigative Proposal and Public Interest

Before substantial investigative work begins, the responsible editor should establish the central public-interest question, principal claims requiring examination, information already known, matters still unverified, likely sources and documents, potentially interested parties, foreseeable conflicts, privacy and safety concerns, relevant jurisdictions, legal considerations, required expertise, proposed reporting methods, right-of-reply process, source-protection requirements, publication threshold, possible outcomes if the evidence remains incomplete and post-publication monitoring needs.

An investigation should not begin with a predetermined instruction to prove a particular conclusion. A working hypothesis can guide research, but the reporting process must allow evidence to change, qualify or disprove that hypothesis.

Public interest may include exposing serious wrongdoing, protecting health or safety, identifying abuse of authority, documenting deception affecting consumers or citizens, revealing rights violations, explaining systemic failures, examining misuse of funds, identifying concealed conflicts, protecting vulnerable people, exposing environmental harm, examining conduct affecting democratic participation or correcting a materially inaccurate public record. Public interest is not the same as curiosity, embarrassment, commercial competition, personal retaliation or the possibility of attracting traffic.

Evidence Plan and Claim-by-Claim Review

Each central claim should have a documented evidence trail identifying the precise proposition being made, available primary evidence, corroborating material, source interests, documentary gaps, alternative explanations, people who may dispute the claim, legal status, technical limitations, appropriate level of certainty and wording justified by the evidence.

A large number of weak sources does not necessarily create strong evidence. Sources should not be treated as independent when they ultimately rely on the same originating person, organization, document or unverified claim.

High-risk investigations should use a claim-by-claim worksheet or an equivalent review record. For each material claim, the record should identify the proposed wording, claim type, supporting documents, source identity, basis of knowledge, corroboration, conflicts, date, jurisdiction, calculation or technical method, right-of-reply requirements, privacy considerations, legal implications, uncertainty, final editorial decision and responsible reviewer. Claims should be qualified, rewritten, removed or withheld when the available evidence does not justify the proposed level of certainty.

Documents and Leaked Material

Investigative documents may include contracts, correspondence, public filings, court records, regulatory records, financial statements, invoices, meeting records, policies, technical logs, datasets, reports, screenshots, source code, audit records, identity records and recordings. Such material should be assessed for authenticity, origin, date, authorship, version, completeness, signatures, metadata, internal consistency, possible manipulation, missing attachments, context, legal restrictions, privacy, security and the source’s potential motive.

A document can establish that a statement was recorded or made without establishing that the statement itself is true. Draft documents should not be presented as final decisions, while unsigned, incomplete or undated records require additional caution.

Leaked material may contain information of substantial public importance, but its confidential or leaked status does not establish authenticity or justify publishing everything contained within it. DesiAI considers how the material was obtained, whether portions are missing, whether it may have been altered, whether the source selected material strategically, whether uninvolved individuals could be exposed, whether credentials or security information are included, whether the public-interest purpose can be achieved through summary or redaction and whether independent confirmation is available.

Complete databases, identity records, passwords, private keys, medical records, intimate material and unrelated personal information will not be published simply because they appear within leaked material.

Data and Financial Analysis

Investigative data work should document the source, collection method, included and excluded fields, missing values, duplicates, date range, jurisdiction, definitions, normalization, calculations, assumptions, currency conversion, inflation treatment, tax treatment where relevant, sampling, error rates, limitations and reproducibility.

Financial reconstruction distinguishes payment from accounting classification, revenue from profit, gross from net, booked amounts from actual cash movement, company funds from personal funds, loans from income, estimates from audited figures and expected amounts from amounts actually received. Allegations are distinguished from established findings.

Where records are incomplete, the limitation should be stated clearly. A reconstruction should not be described as an audit unless it was actually conducted as an audit by an appropriately qualified party.

Technical Testing

Technical investigations involving software, platforms, cybersecurity, privacy or artificial intelligence must be lawful, proportionate, documented and limited to what is necessary to establish the relevant issue. Testing should minimize potential harm, be conducted by someone with appropriate competence and be reviewed for security and privacy implications.

A test result should identify relevant details such as device, software version, date, configuration, account type, jurisdiction, network conditions, procedure, sample and known limitations, along with whether the result can reasonably be generalized.

Testing a single account, device or configuration does not automatically establish universal behavior. Security vulnerabilities are handled under the Security and Vulnerability Disclosure Policy and should not be exploited beyond what is reasonably necessary to demonstrate the issue safely.

Human Sources and Anonymous Sources

Investigative sources are assessed according to their direct knowledge, position, access, timing, consistency, documentary support, motivation, personal disputes, employment status, financial interests, legal exposure, political interests, reliability history and potential risk of retaliation.

A source with a grievance may still provide accurate information, but the grievance is relevant when assessing motive and corroboration. A senior title does not necessarily establish direct knowledge, and a former employee may have useful knowledge of historical practices without knowing how the organization currently operates.

Anonymous sources are governed by the Anonymous and Confidential Sources Policy. The responsible editor should normally know the identity of a source who is central to an investigation, except in an extraordinary circumstance approved at the highest appropriate editorial level.

Anonymous-source claims should ordinarily be corroborated through independent sources, documents, recordings, direct observation, data, public records or technical evidence. Anonymity must not be used to publish unsupported accusations, speculation or personal attacks.

Undercover and Deceptive Methods

Deception, concealed identity, covert recording or undercover reporting may be considered only when the information sought is of significant public importance, there is a reasonable basis to suspect serious misconduct, less intrusive methods have failed or are unlikely to succeed, the method is proportionate, foreseeable harm has been assessed, applicable legal requirements have been considered, advance approval has been obtained, the activity is narrowly limited and the resulting evidence can be independently evaluated.

Undercover methods are not used for entertainment, humiliation of private individuals, ordinary consumer testing, creating misconduct that would not otherwise occur, collecting irrelevant personal information, unlawfully accessing systems or premises or pursuing a purpose whose expected public value is minor.

Where doing so does not create an overriding safety or legal risk, the use of deceptive or undercover methods should be disclosed in the published work.

Recording and Evidence Integrity

Audio and video recording must comply with applicable law and relevant ethical standards. Even where covert recording is legally permitted, its necessity and proportionality should be assessed before use.

Recordings should be stored securely and should not be edited in a way that changes their meaning. Transcripts should identify uncertainty, interruptions and translation issues where relevant. A short excerpt should not be presented in a manner that creates an impression contradicted by the complete recording.

Fairness and Right of Reply

People and organizations facing serious criticism or allegations should receive a meaningful opportunity to respond before publication unless exceptional urgency, safety or another compelling reason makes this impracticable.

A request for comment should identify the principal allegations or factual claims, relevant dates, documents requiring explanation, response deadline, applicable time zone, contact person and intended publication timing where appropriate. DesiAI does not have to disclose confidential sources, unpublished evidence that could create significant risk or complete drafts simply to facilitate a response.

The response period should reflect urgency, complexity, number of allegations, availability, language, time zone and the need to review relevant records. Responses should be considered and represented fairly. A refusal or failure to respond is not treated as proof that an allegation is true.

Limited pre-publication review may be offered for technical accuracy, specialist terminology, translated quotations, safety-sensitive descriptions, complex chronology or confirmation of personal details. Such review does not give the subject control over the conclusion, headline, sources, structure or publication schedule.

Privacy and Harm Minimization

Investigative reporting can uncover information about people who have a reasonable expectation of privacy. Publication should therefore be limited to information that is relevant and proportionate to the public-interest purpose.

Editors should consider whether the person is a public or private individual, their role in the matter, seriousness of the conduct, age, vulnerability, risk of retaliation, impact on family members, location, medical or intimate information, long-term search consequences, whether anonymization would preserve the public value and whether identifying details are genuinely necessary.

Private information should not be added simply to make an investigation more dramatic. Uninvolved relatives, junior employees, witnesses and incidental individuals should not be exposed without a defensible reason. Children and vulnerable people receive heightened protection, including careful consideration of consent, future digital consequences, safety, retaliation and identification through combinations of otherwise harmless details.

Language, Legal Review and Security

Investigative language must reflect the strength and status of the evidence. Articles should distinguish between allegations, documented events, witness accounts, internal findings, regulatory decisions, court findings, criminal charges, civil claims, admissions, denials, reasonable inferences and editorial conclusions.

Terms such as fraud, corruption, theft, criminal, illegal, scam, abuse and deception should be used only when supported by the available evidence and appropriate to the legal and factual circumstances. Headlines must not create a stronger allegation than the article itself establishes.

Qualified legal review should be considered for serious allegations, privacy issues, confidential information, children, criminal proceedings, court restrictions, leaked material, undercover reporting, copyright, data protection, unauthorized access, employment confidentiality and complex cross-border matters. Legal review assists with risk assessment but does not replace editorial judgment.

Investigative records should be secured according to their sensitivity. Appropriate controls may include restricted access, encrypted storage, separate source-identity records, secure communications, metadata removal, controlled backups, limited printing, secure file transfer, device protection and appropriate deletion practices. No security measure should be represented as completely risk-free.

Final Approval and Methodology Disclosure

Before publication, the responsible editor should confirm that the Investigation label is justified, central claims have documented evidence trails, significant contrary evidence was considered, the right-of-reply process was reasonable, anonymity decisions were approved, conflicts were addressed and relevant media was verified.

The review should also consider privacy and harm, legal issues, headline proportionality, commercial interests, methodology disclosure, post-publication monitoring and whether the final published page matches the approved version. High-risk investigations should receive final approval from the Publisher or another appropriately authorized senior editor.

Where appropriate, an investigation should include a Sources and Methodology section describing principal evidence, the number and type of interviews, period covered, documents reviewed, data methodology, technical testing, requests for comment, limitations, use of anonymous sources, material supplied by interested parties and unresolved uncertainty. Confidential sources and sensitive reporting methods should remain protected.

Publication Timing, Collaboration and Post-Publication Duty

Publication timing should be based on reporting readiness, public interest, safety and legal considerations. DesiAI should not publish prematurely merely to beat another publication, meet a search deadline, satisfy a client, influence a developing event without adequate evidence, prevent a meaningful response or capitalize on temporary public attention.

Delay may be appropriate while evidence, verification, responses, safety assessments, legal review or court restrictions remain unresolved. However, delay should not be used indefinitely to shield powerful interests from accurate scrutiny.

Collaborations with other publications, researchers or organizations should establish expectations around editorial control, source protection, evidence standards, legal review, timing, attribution, copyright, correction responsibility, security, commercial interests and dispute resolution. A partner’s contribution should be disclosed when it materially shaped the investigation.

Following publication, DesiAI should remain attentive to factual challenges, new documents, responses, legal developments, source safety, security concerns, misleading excerpts, correction requests and evidence that a source provided materially misleading information.

An investigation should be corrected, clarified, updated, noted or retracted when the evidence requires it. The time, resources or reputation invested in an investigation do not justify defending an article that is no longer supported by reliable evidence.

Contact

Specific factual concerns regarding an investigation may be sent to corrections@deshiai.in. Formal complaints, rights of reply and concerns about investigative conduct may be sent to complaints@deshiai.in.