What Is the Ethical Issue?
Artificial intelligence tools — large language models, image generators, audio synthesis, and video deepfake systems — are now integrated into the workflows of many UK newsrooms. The ethical obligations these tools create are still being defined by regulators, unions, and professional bodies. What is clear is that AI does not change the journalist’s fundamental accountability: for accuracy, for fairness, for transparency, and for the impact of their work on subjects and audiences.
The NUJ has published AI guidance emphasising that AI must be a tool that enhances journalism, not one that replaces journalist oversight. IPSO’s position is that the Editors’ Code applies regardless of how content is produced: a story that is inaccurate because an AI hallucinated a fact is still a Clause 1 breach. The journalist and publisher are responsible, not the AI system.
The ethical framework for AI in journalism centres on three areas: transparency (being honest with readers about how content was produced); verification (not publishing AI output without human fact-checking); and accountability (the journalist remains responsible for everything that carries their by-line or their publication’s name).
When This Matters in Practice
AI-drafted news summaries
A newsroom uses an LLM to draft initial summaries of press releases or public documents. The ethical obligation: a journalist must read and verify the output before publication. Factual claims in the AI draft must be checked against the primary source. Headlines and summaries must accurately represent the body of the story.
AI-generated images
An AI image generator is used to create an illustration for a feature article. The ethical obligation: disclose clearly that the image is AI-generated. Do not use synthetic images of real identifiable individuals without their consent, as this can amount to misrepresentation. Archive the prompt used to generate the image.
Deepfake audio and video
Voice cloning or video synthesis technologies can be used to create synthetic interviews, dramatisations, or illustrations. Any such content must be clearly labelled as synthetic. Using deepfake technology to put words into a real person's mouth without their consent is unethical and potentially unlawful under the Online Safety Act 2023.
AI as translation aid
AI translation tools are used to translate source documents from another language. The ethical obligation: the translation must be verified by a competent speaker of the source language before publication, especially for anything that will be quoted directly. Machine translation errors in legal, scientific, or political contexts can be seriously misleading.
AI for research and verification
A journalist uses an AI tool to search and summarise published research on a topic. The ethical obligation: all AI-generated summaries must be verified against the original source documents. AI tools regularly misrepresent academic papers, fabricate citations, and confuse study populations or findings. Never cite a source you have not read.
Red Flags
- ⚠Publishing AI-generated content without human review and fact-checking
- ⚠Using an AI-generated quote attributed to a real person without verification
- ⚠Using AI output as a primary source for factual claims
- ⚠Generating synthetic images of real individuals without disclosure or consent
- ⚠Using audio or video synthesis to create fake interviews or statements
- ⚠Failing to disclose that content was primarily AI-generated to readers
- ⚠Using AI translation of critical documents without expert verification
- ⚠Relying on AI to check legal or scientific claims without reading the primary sources
- ⚠Archiving AI-generated content to a publication's record without noting its AI origin
- ⚠Using AI tools that were trained on copyrighted journalism content without assessing the legal risk
AI Content Pre-Publication Checklist
Generate Your AI Disclosure Statement
Use the AI Disclosure Checklist to decide whether disclosure is required and, if so, to generate appropriate disclosure language for your specific use case.
Open AI Disclosure Checklist →Common Mistakes
- ›Treating AI output as verified fact: AI language models generate fluent, confident text that can be completely wrong. Statistics, case citations, study findings, biographical details, and dates are all high-risk areas for hallucination. Every AI-generated factual claim must be verified against a source you have actually consulted.
- ›No disclosure for AI-generated images: An AI-generated photograph-style image that is not disclosed can mislead readers into believing something was photographed that was not. This is a form of misrepresentation. The BBC, Reuters, and AP all require disclosure of AI-generated imagery.
- ›Assuming AI tools are legally neutral: AI tools trained on copyrighted content may generate outputs that infringe copyright. AI-generated text that closely resembles a specific author's work may constitute infringement. AI images based on real individuals' likenesses raise personality rights issues. Legal assessment is required before widespread deployment.
- ›Using AI for editorial decisions without oversight: AI tools that are used to select stories, write headlines, or determine placement are making editorial decisions. These decisions affect what the public reads and how they understand the world. They require editorial oversight, testing for bias, and transparency with audiences.
- ›Archive integrity: AI-generated content, once published, becomes part of a publication's archive. Future readers and researchers using the archive may not know the piece was AI-generated. Metadata, disclosure statements, and archive notes should record the use of AI in production at the point of publication.
UK Newsroom AI Policies: Examples
- BBC: Prohibits using AI to generate or significantly alter editorial content, including text, audio, and video, without editorial oversight. AI-generated images must be disclosed. Editorial use of AI requires sign-off under the BBC Editorial Guidelines.
- The Guardian: Has published a policy allowing limited AI use as a research and production tool with mandatory human oversight. AI-generated content published with the Guardian’s name requires editorial review and disclosure.
- Reuters: Issued AI usage guidelines requiring human editorial oversight of all AI-assisted content. Reuters’ AI principles include not using AI to replace journalists in core reporting functions without disclosure.
- Financial Times: Has a newsroom-wide AI policy requiring that AI tools be approved before use, that all outputs be verified, and that any use in published content be reviewed by an editor.