Gains:
- Ability to apply the principles of accuracy, transparency, accountability, independence and non-maleficence in the use of artificial intelligence
- Ability to meaningfully disclose the use of AI to the reader and apply the standard of using synthetic visual/audio only with clear labeling
- Understand that responsibility cannot be delegated to the agent and that a written newsroom AI policy ensures consistency and accountability
Using artificial intelligence in journalism is as much a matter of trust as it is a technical issue. The reader has the right to know how a news story is produced; sources must know how information about them is processed; The profession must uphold the principles of integrity and independence. In this unit, we establish the ethical framework of AI use: accuracy (verify), transparency (clearly declaring use), accountability (taking responsibility), independence and non-maleficence. The basic principle is simple: AI is a tool; Journalistic principles do not change when the medium changes. Whether AI is used or not, the journalist is responsible for the accuracy and ethical consequences of each sentence. This unit is the moral framework that sits on top of all previous skills.
Five ethical principles and their counterparts in AI
1) Accuracy. AI is fluid but fallible. The principle of accuracy dictates that every AI output be verified from the primary source (Unit 1-7). "AI wrote it" is no excuse.
2) Transparency. The reader should know how AI is involved in the process. This is not a footnote in every sentence; It is a meaningful statement. For example, whether a news story was drafted entirely with AI, whether an image was produced with AI, whether a translation was made with a machine - these are explained to the reader.
3) Accountability. A human being stands behind everything published. If there is a mistake, it cannot be said that "the vehicle made it"; The responsibility for correction and refutation belongs to humans.
4) Independence. AI tools may carry certain frameworks, biases and commercial interests. The journalist does not make the frame of the medium the frame of the news; maintains editorial independence.
5) Do no harm. Content produced with AI may harm real persons (false attribution, violation of privacy, damaging reputation with synthetic visuals). The public benefit-harm balance is always established by human judgment.
Caution: Transparency is not “no hiding, no problem.” The reader would be misled if he did not know that an image was produced by AI or a text was translated by AI. Failure to declare usage is an active breach of trust.
AI-generated visuals and audio: special attention
Using an image/sound produced by AI in the news is a special ethical area. A synthetic image presented as representing a real event amounts to disinformation. Principle: AI-generated visual/audio is used only if it is clearly labeled and does not imitate the real event (e.g. obviously a representative drawing). Synthetic media that give the impression of reality are not included in the news.
Newsroom AI policy: what does it cover?
Having an organization have a written AI use policy ensures both consistency and accountability. A good policy clarifies: in which tasks AI is allowed and in which it is prohibited; which vehicles are approved; sensitive data rules; transparency/labelling standard; verification obligation; chain of responsibility. AI can also help draft such a policy — but approval is given by editorial management.
three mini cases
Case 1 — Unlabeled synthetic image. A publication used a "crime scene" image produced with AI as if it were real in a news article. Readers noticed and reacted; The publication lost trust and issued an apology. If the image had been labeled "representative, produced with AI", the problem would not have arisen.
Case 2 — Reliance on transparent representation. One publication wrote clearly at the end of the news that it used AI in a data analysis and how it verified it (“Analysis scanned with AI support, all numbers confirmed from official record”). Reader confidence increased; Interest in similar news increased. Transparency has become a source of trust, not weakness.
Case 3 — "The vehicle did it" defense collapsed. A news report made a false accusation about AI. "That's how the AI summarized it," the reporter said; but due to the principle of accountability, the responsibility was on the air. The denial was published, and the newsroom wrote a policy requiring the fact-checking step. Lesson: responsibility cannot be delegated.
Copiable templates
1) Draft transparency statement:
In a news story, I used AI in the following tasks: [e.g. transcription, first draft, data screening, translation]. I have verified all facts from the primary source. Write an honest and concise "AI use statement" sentence aimed at the reader; do not exaggerate, disguise, or use unnecessary technical language. Usages: [here]
2) Ethical risk screening:
Scan the news draft below for ethics. Mark the following: (1) unsubstantiated attribution, (2) privacy risk, (3) potentially damaging language, (4) unlabeled use of AI-generated content, (5) one-sided/biased framing. Only a risk report is issued. Draft: [here]
3) AI usage policy framework:
Create a draft AI usage policy for a newsroom. Headings: allowed/prohibited tasks, approved tools, sensitive data rules, transparency/labeling, verification obligation, chain of responsibility. Write 2-3 sample rules under each heading. This is a draft; Approval is with the management.
4) Visual/audio labeling decision:
I am considering using the following AI-generated image/audio in the news: [description]. Consider whether it gives the impression of authenticity, the risk of disinformation, and the appropriate label. Write the recommended label text if available; if not, explain the rationale.
Weak prompt / Strong prompt
Weak prompt: "Is it ethical to publish this news?"
Result: AI gives a vague "usually OK" answer; The real risk points remain invisible, the decision becomes unwarranted.
Strong prompt: "Scan this news draft for ethics and flag concrete risks under the following headings: unverified attribution, privacy, harm, unlabeled AI content, biased framing. The decision is mine; you just issue a risk report."
The difference: Powerful prompt turns the AI into a concrete risk scanner, leaving the decision up to the journalist and producing an auditable list rather than vague confirmation.
comparison chart
principle
Risk in AI
Application
accuracy
hallucination
Confirmation from primary source
transparency
confidential use
Meaningful statement/label
accountability
"The vehicle did it" excuse
Responsibility lies with people
independence
Instrument bias
Maintain editorial frame
do no harm
Imputation/privacy
Public benefit-harm balance
Common mistakes
- Transferring responsibility to the vehicle. The "AI wrote/summarized" defense is invalid; The account belongs to the human.
- Hiding usage. Not declaring the image/translation/text produced by AI is a breach of trust.
- Presenting the synthetic image as real. Unlabeled AI imagery opens the door to disinformation.
- Adopting the framework of the tool. Bringing AI's bias into the news frame without questioning it.
- Working without politics. Without written rules, everyone acting differently creates inconsistency and risk.
How to implement transparency? Measured and meaningful statement
While the principle of transparency sounds simple, its implementation requires finesse: it is neither possible nor meaningful to write "this was written with AI" at the end of every sentence; It overwhelms the reader with a flood of information and makes the really important statements invisible. Good transparency is measured and meaningful: it describes uses of AI that will influence the reader's evaluation of the news. An image being produced by AI, a translation being made by machine and a quote based on machine translation, a news being drafted largely by AI – these are things the reader needs to know. In contrast, simply checking spelling or brainstorming a title may be considered routine enough to not require a separate statement. The measure is “does this information change the reader's confidence in the news?” It is found by asking.
The second dimension of transparency is towards sources. When talking to a source, being honest about how what they say will be processed (e.g. deciphered by an AI tool) preserves the trust relationship; Particularly sensitive sources have the right to know what means their data will pass through. The third dimension is internal: a reporter must make it clear to his editor and the fact-checking desk where he is using AI in a story so that the editorial chain can control what is human and what is machine. Transparency is not an admission of weakness; It is the profession's commitment to honesty towards the reader, the source and the institution.
In summary
AI is a tool; Journalism's principles of accuracy, transparency, accountability, independence and non-maleficence do not change when the medium changes. Each output is verified from the primary source; The use of AI is meaningfully declared to the reader; synthetic visual/audio is used only with clear labeling and without imitating reality; The responsibility always lies with the person. A written newsroom AI policy ensures consistency and accountability.
Application task
Apply the “Ethical risk screening” prompt to your draft of a story; Really evaluate each risk that arises and write down how you will eliminate it. Then, with the “Draft transparency statement” prompt, prepare an honest AI use statement to include in your story and discuss its impact on reader trust in one paragraph.
checklist
- [ ] Each AI verifies its output from the primary source; I take responsibility.
- [ ] I expressly explain the use of AI to the reader.
- [ ] I use synthetic visual/audio only with clear labeling and without imitating the real thing.
- [ ] I question the bias of the medium and maintain editorial independence.
- [ ] I comply with or advocate for the creation of a written AI use policy.