AI Accountability In Newsrooms: Who Answers When AI Gets A Story Wrong?
My position is that when AI gets a story wrong, the newsroom that published it answers for it, and inside that newsroom a named person should answer. Not the model, not the software vendor, not the freelancer who handed it over. But there is a catch that most policy documents skip: accountability only means something if that person has the time, the authority and the records to do the job. This piece looks at three cases where blame went elsewhere, what the rules already say, and the strongest objection to my view. It is a question of AI accountability in newsrooms, and who is responsible for AI errors when the chain of hands is long.
What AI accountability in newsrooms should mean
In plain terms, it means three things. First, a named person owns each AI use and each AI-assisted story. Second, that person has the time and authority to say no. Third, there is a record showing what the AI did, what was checked and who approved publication.
The Partnership on AI, a nonprofit that publishes guidance for newsrooms, makes a similar point: a single person should be identified as responsible for decision-making and oversight of an AI tool, so that if the tool makes mistakes, it is clear who to contact to make changes. My argument goes one step further. Naming a person is necessary and not sufficient.
Three cases where blame went elsewhere
The Chicago Sun-Times, 2025. A summer supplement included a reading list of 15 books, of which 10 did not exist. The Sun-Times said the section was licensed content that was not created or approved by its newsroom. The freelancer who wrote it for the syndicator King Features said he had republished what an AI tool produced, and King Features cut ties with him. The Philadelphia Inquirer also ran the supplement. One report noted it was not clear who at King Features had responsibility for editing the freelancer’s material.
Read that sequence again. Each party can explain how the error happened. But the readers saw a Sun-Times page. The newspaper later said it would identify third-party content in print and review its relationships with contractors, which is an admission that accountability sits with the publisher.
Wired and Business Insider, 2025. A freelancer using the name Margaux Blanchard appears not to exist, yet got articles into several outlets. Wired said it became suspicious when the writer could not provide enough information to be entered into its payment system. An editor then ran the story through two commercial AI-detection tools, and Wired’s own account says the piece had slipped past human review and those tools before editors found fabricated details and retracted it. Business Insider removed the articles and, after a wider review, pulled dozens of essays tied to questionable bylines; reports vary between at least 34 and 38. It said it had strengthened its verification protocols.
Notice who answered. Neither outlet blamed the AI. They answered for their own commissioning and verification.
Air Canada, 2024. This is not a news case, but it is the clearest legal statement I have found. A tribunal in British Columbia, in Moffatt v. Air Canada, rejected the airline’s argument that its chatbot was a separate legal entity responsible for its own actions. It held the airline responsible for all the information on its website, whether it came from a static page or a chatbot, and awarded the customer a small sum. It is a small-claims decision in a consumer setting, so it does not bind courts handling news or defamation disputes. But the principle fits: you cannot hand your accountability to the tool you deployed.
What the rules already say
AP. The Associated Press’s updated AI standards, released in July 2026, say editorial judgment, verification and accountability remain the responsibility of AP journalists.
The EU AI Act. Article 50 became enforceable on August 2, 2026. For AI-generated text published to inform the public on matters of public interest, deployers must disclose that it was AI-generated, unless the content has undergone human review and a natural or legal person holds editorial responsibility. In other words, the law itself asks who has editorial responsibility. One legal guide adds that publishers relying on that exemption should be ready to show logs demonstrating the extent and nature of human involvement. This is not legal advice, and whether the Act applies to you depends on where you operate.
Practitioner guidance. One practitioner guide says plainly that the newsroom is responsible, not the company that made the AI tool, noting that software agreements usually rule out the AI company being blamed for mistakes or defamation. Read your own vendor contracts to see whether that holds for you.
The strongest objection to my view
The best objection comes from a researcher interviewed by UVA Today, who warns about humans becoming perpetual scapegoats. The example is a radiologist who must sign off on every AI diagnosis but is not given time to genuinely second-guess it. They are accountable on paper without having any meaningful say. The same researcher argues that the organizations building these systems should also bear accountability when they fail.
I agree with most of this, and it changes how I state my own position:
Accountability without time is a trap. The Partnership on AI notes that CNET editors reviewed AI content, according to The Verge, but often months after publishing, which limited the value of the review. A named reviewer who has no time is a scapegoat in waiting.
Vendors and syndicators should share the risk. Contracts can require disclosure of AI use, warranties on accuracy and cooperation after an incident. Publisher accountability to readers does not stop you from pushing risk back to the party who caused it.
Small newsrooms face real limits. Business Insider itself said limited resources at smaller outlets make due diligence increasingly challenging. The answer is not to lower the bar but to narrow the use: fewer AI tasks, tighter review.
A scholarly name for the problem is the “many hands” problem: when many people and systems contribute, it becomes hard to say who is responsible, and blame tends to land on whoever is nearest. That is precisely why the owner has to be named before something goes wrong.
What a real accountability chain looks like
List allowed and disallowed AI uses. For example, transcription and translation allowed, unchecked AI-written copy not.
Name an owner for each use, and a named editor for each AI-assisted story.
Give that owner time and the right to stop publication, without penalty.
Keep a record of what the AI did, what was checked and who approved, using an AI audit trail.
Treat third-party content as unreviewed until your own gate has checked it, as the Sun-Times case shows.
Check vendor and freelancer contracts for AI-use disclosure and accuracy terms.
Disclose AI use according to your policy and local rules.
Plan the correction path in your corrections policy so errors are fixed fast and openly.
Review incidents and update the chain afterward.
For the approval step itself, see our piece on the human gate before AI publishes.
Where NewsBolts fits
At NewsBolts, we build on one principle: AI helps with research, drafting and publishing tasks, and a named person decides what goes live. A workflow that records who approved what is how that principle turns into accountability you can show. You can read more in our AI newsroom operating system guide.
FAQ
Who is responsible when AI gets a news story wrong? In my view, the newsroom that published it, and a named editor inside it. Vendors, freelancers and syndicators may share legal or contractual responsibility, but readers hold the publisher accountable.
Can a newsroom blame the AI tool? Not credibly. In the Air Canada case, a tribunal rejected the idea that a chatbot was a separate entity responsible for itself, and AP’s standards keep accountability with its journalists.
What does the EU AI Act say about editorial responsibility? Article 50(4) requires disclosure of AI-generated public-interest text unless it has undergone human review and a person or organization holds editorial responsibility. Check how it applies to you with a lawyer.
How can a small newsroom do this without much staff? Narrow the AI uses you allow, name one owner, keep a short log for each AI-assisted story and treat third-party content as unreviewed until you have checked it.
Conclusion
When AI gets a story wrong, “the tool did it” is not an answer, and “the freelancer did it” explains the error but does not remove the publisher’s accountability. Name the person, give them time and authority, and keep a record. Then, when something does go wrong, you can say who decided, what they checked and what changes next.
Comments