Arab News has embraced artificial intelligence at scale, including deploying CAMB.AI to translate its journalism into 50 languages, but Editor-in-Chief Faisal Abbas says the technology has been deliberately kept away from the core functions of journalism. In an interview, Abbas says the newsroom uses AI for translation, research, idea generation, document summarisation and SEO-friendly headline suggestions, while journalists retain responsibility for reporting, verification and editorial judgment. He also argues that AI’s biggest lesson for media and brands is the need for a strict human-in-the-loop system, warning that agencies using AI are increasingly facing client expectations for shorter timelines and lower fees.
Arab News went all-in on CAMB.AI while most regional newsrooms are still running pilots. What did that first year of real production use teach you that the pitch deck didn’t?
The pitch deck sells speed, seamless scale, and infinite translation. The reality of full-scale production teaches lessons about trust, workflow integration, editorial control, and audience behavior. It is basically a workflow transformation regardless of the technology, which has to start with an internal mindset and dedication towards achieving an overall objective.
Year one also taught us that localized content needs tailored distribution pipelines. It was not enough that we pressed the button and launched, we had to tell everyone (for 50 languages) about it, for reach and usage and to benefit from the service, as we developed a weekly ongoing campaign with the help and collaboration of country’s Ambassadors in Saudi Arabia.
You’ve said AI should empower journalists, not replace them. Where has that line actually been tested? Any moment the tech pushed further than you were comfortable with?
The line between empowerment and replacement gets tested the moment AI moves from handling mechanics (research, translation, summaries etc.) to attempting judgment. Hence why journalist roles remain important to the operation as Ai still requires human interaction and checking. Which brings us back to the initial statement of empowering journalists and not replacing them
50 languages is the headline number. What breaks first, accuracy, nuance, or trust, when you scale translation AI across dialects as different as Gulf and Levantine Arabic?
We have not added Arabic as part of the 50 languages as we are part of a group that has several titles in Arabic and did not see the need for overlapping. However, when we do translate certain articles manually to Arabic using AI for various other channel distributions, it is always best to use the standard written professional Arabic dialect and not get into spoken ones from various countries and areas.*
Newsrooms are supposed to be the ones fact-checking everyone else. What’s your internal verification process for AI-translated or AI-assisted content before it publishes?
We have 2 tracks for this question:
One, 50 languages translation utilizing CAMB.AI technology: Ahead of launching we ran several months of testing and verifying the efficiency of the translation for each language through our global connections with contributors, correspondents and writers. This proved to us that the technology CAMB.AI is using is very advanced and efficient, with an overall average of 97%-98% efficiency. Also, the source is always the English text that anyone can refer back to for checking.
Two, as for manually AI-translated content, we have a full translation and also a copy editing department that checks and verifies the content compared to the original source before proceeding with publishing.
Translation is one use case. Where else has AI actually embedded itself in the Arab News newsroom day-to-day, story generation, research, headline testing, audience data, and where have you deliberately kept it out?
Other than translation, we use AI for various activities throughout the newsroom and departments. This includes idea generation, research, summarizing large documents – tasks that used to take hours from our reporters and editorial time ahead of writing, now can be done in minutes, allowing our editorial team to spend their time where they need to, and be more efficient.
We have also just launched our new in-house CMS tool “Maestro” that embeds Ai at every step where it can help our editorial team – an example would be, automatically suggesting SEO-friendly headlines.
We have deliberately kept AI out of actual story generation and editorial judgment. AI does not write news stories or articles, it does not conduct interviews, does not publish breaking news.
Given OpenAI’s and other major players’ public stumbles this year, has that shaken confidence in the AI vendors regional newsrooms and brands are building on, including your own partners?
Not at all – our dependency on AI usage when it comes to the newsroom is mostly backend as explained, and the public stumbles of broad consumer AI platforms have vindicated our approach; relying on our editorial team and journalists for the core of the operations.
Also it proves that in media you don’t bet on general-purpose hype, you partner with specialized platforms built for trust, security, and enterprise uptime.
You cover the marketing industry as closely as you practice journalism. When brands or agencies get AI wrong publicly, what does that look like from a newsroom’s vantage point, and what do they usually get wrong about how the press will treat it? From where you sit, are advertising agencies actually building AI into daily production, or is most of what you’re seeing in pitches and press releases still theater?
When a brand blunders publicly with AI—whether it’s a “soulless” fully AI-generated commercial, hallucinated product copy, or a tone-deaf automated campaign—the newsroom views it through a very simple lens: a breakdown in human accountability. From a newsroom vantage point, an AI blunder isn’t a software bug; it’s a decision by leadership to prioritize speed and margin over quality control, ethics, and cultural intelligence.
Journalists don’t judge the tool; they judge the lack of a human-in-the-loop. When an agency releases an ad with distorted AI artifacts or insensitive context, the story isn’t “Look what the AI did”—it’s “Look what this brand was willing to sign off on.”
What we noticed is that most of these remain one-off novelty stunts. They win short-term tech coverage but rarely represent a scalable, repeatable agency business model.
Ai helps agencies when actively using it for instant storyboarding, dynamic asset variation at scale (formatting one master ad into dozens of ratio/language iterations), rapid pre-visualization, and processing consumer research data.
However, the real tension in daily production isn’t whether agencies use it—it’s that clients now expect a 30% reduction in timelines and fees because they know the agency is using AI to handle the heavy lifting.
If you were advising a CMO nervous about AI-generated content facing real customers, what’s the one guardrail you’d insist on, based on what Arab News had to build internally first?
If forced to pick the single non-negotiable guardrail based on Arab News’ internal journey, it is establishing a strict “Human-in-the-Loop” Verification Protocol (or Editorial Firewalls) before any piece of content reaches a live audience.
At Arab News, AI generates speed, scale, and draft variants, but it never gets publishing privileges. For a brand facing real customers, this means drawing a firm operational boundary between content generation and content release.
In newsrooms, an unverified fact or incorrect terminology ruins trust instantly. For a brand, hallucinated product specs, incorrect pricing, or compliance errors in customer-facing copy create immediate legal and financial exposure.
—
