Guide

Covering AI responsibly: an ethics checklist for tech news in 2026

AI is now one of the biggest beats in tech news, and one of the easiest to get wrong. Launch videos are slick, benchmark charts are selective, and the pressure to publish first is intense. Meanwhile audiences are making real decisions, about jobs, money and privacy, based on what they read and watch.

This checklist is for anyone covering AI: reporters, YouTubers, newsletter writers and podcasters. It updates older guidance for 2026, now that the EU AI Act is in force and AI-generated media is everywhere.

1. Accuracy: separate claims from facts

The IPSO Editors' Code of Practice opens with accuracy: the press must take care not to publish inaccurate, misleading or distorted information or images, including headlines not supported by the text, and significant inaccuracies must be corrected promptly and with due prominence. Even if you aren't an IPSO member, that is a good standard to hold yourself to.

For AI coverage, that means:

Attribute every capability claim. "The company says" is different from "the model can".
Distinguish demos from products. Was it a live, unedited demo, a pre-recorded video, or something you used yourself?
Read benchmarks critically. Who ran the test, on what data, compared with which models? Was the comparison chosen by the vendor?
Go to primary sources. Read the technical report, model card or regulatory filing, not just the press release or someone else's thread.
Make headlines match the story. A thumbnail that says "AI replaces doctors" over a video about a narrow diagnostic study is misleading.

2. Bias and fairness: ask who is affected

AI systems can perform differently for different groups of people, depending on how they were trained and used. When covering a system that makes or informs decisions about people, such as hiring, lending, policing or healthcare, ask:

Has it been tested on people like those it will be used on?
Who audited it, and are the results public?
What happens when it gets something wrong, and can people challenge the result?

The ICO's guidance on AI and data protection covers fairness, bias and discrimination, transparency, and statistical accuracy. It is a useful source of questions to put to companies deploying AI on UK residents' data.

Seek out voices beyond the company and its investors: affected users, independent researchers and civil society groups.

3. Transparency: disclose your own AI use

If you use AI to research, transcribe, translate, write scripts or generate images, tell your audience. A short, standing note on how you use AI, plus a label on any AI-generated visuals or audio, costs little and builds trust.

In the EU this is increasingly a legal matter as well as an ethical one. According to the European Commission's AI Act overview, the Act entered into force on 1 August 2024 and became applicable on 2 August 2026, with some parts, such as the bans on prohibited practices, applying earlier. Article 50 sets transparency obligations: providers must ensure AI-generated content is marked in a machine-readable format and detectable as artificially generated or manipulated, and deployers who create deepfakes must disclose that the content has been artificially generated or manipulated. There are narrower obligations for evidently artistic, satirical or fictional work, so read the text if that applies to you.

4. Verification: assume media can be synthetic

Treat viral clips of public figures with suspicion until you can trace them to an original source.
Check metadata and look for provenance labels, but don't rely on their absence to prove anything.
Contact the person or organisation shown before publishing.
If you can't verify something, say so, or don't run it.

5. Privacy: don't feed people's data to tools carelessly

Pasting interview transcripts, leaked documents or source details into a third-party AI tool can expose people who trusted you. Check a tool's data retention and training terms first, and remove identifying details where you can. If you process personal data about UK residents, UK GDPR still applies.

6. Accountability: follow the money and the risk

Disclose conflicts of interest: sponsorships, affiliate links, free access, equity or advisory roles.
Name who benefits from a claim, and who carries the risk if it's wrong.
Revisit big stories. Did the product ship? Did the promised capability hold up? Follow-ups are where the most useful reporting often sits.

7. Corrections: make them easy to find

Mistakes are inevitable in a fast-moving field. Publish corrections clearly on the original piece, pin a correction comment on videos, and note the date. Treat a correction as part of the story, not an embarrassment.

The checklist

Every capability claim is attributed, and independent verification is noted.
Demos, benchmarks and shipped products are clearly distinguished.
Primary sources are read and linked.
Headlines and thumbnails match the content.
Affected people and independent experts have been asked for comment.
Your own AI use is disclosed, and AI-generated media is labelled.
Viral or sensitive media has been traced to its source.
No source or interviewee data has been pasted into tools without checking terms.
Conflicts of interest are disclosed.
There is a visible, dated way to issue corrections.

Frequently asked questions

Do I have to label AI-generated images in my videos?

If you publish in the EU, Article 50 of the AI Act requires deployers who use AI to create deepfakes, meaning realistic content that falsely appears authentic, to disclose that it is artificially generated or manipulated. Labelling AI visuals clearly is good practice everywhere.

What is the most common mistake in AI news coverage?

Repeating vendor claims as fact. Benchmarks, demo videos and launch posts are marketing until someone independent has tested them. Say who made the claim and whether it has been verified.

Does the EU AI Act apply to UK creators?

It can, if your content or AI systems are offered to people in the EU. The UK has its own rules on accuracy and data protection, including the IPSO Editors' Code for member publishers and ICO guidance on AI and personal data.

Sources and further reading

Facts checked on 20 October 2026. Spotted an error? Tell us and we will correct it.