It's hard to trust AIs if you don't trust the companies behind them

The hardest thing for publishers navigating AI and other tech is knowing what the people behind them might do next. The tech ain't the problem.

by Rich Fairbairn

Published: 15:19, 28 August 2025

It's hardly a revolutionary statement, but we all know that trust in media is important. It underscores relations between publishers and audiences, and from a simple business perspective someone is unlikely to become a subscriber or follower if they don't trust you. Trust is what catalyses the relationship to attain and grow value, and you risk it at great peril.

Trust, to media boardrooms, isn't only about the relationship between publishers and audience. Trust equally has to exist between publishing/media firms and the various technologies, both physical and virtual, which hold their future in a vice. There needs to be trust that they work as intended, and also trust that they are being offered fairly - both to protect that vital audience bond, and also to allow staff and leadership to sleep at night while tech firms which move fast and break things have such a hold over their businesses, an unspoken existential stressor amongst modern media workforces. Ultimately, the media has as much right to demand trust in its tech as their audiences do of the media.

It's those collective beliefs which make so many of us increasingly dumbfounded by the recent actions of some of the AI platforms desperately trying to convince us all to become even more reliant on their services and products. 

Distrust built in

OpenAI showed in August that for all its supposed genius, it is a young company still either naive or disingenuous in dealing with real people or enterprise customers. It has a poor reputation in media and publishing circles anyway but it did little to help when the launch of its latest GPT5 model saw the abrupt removal of several older models which were in constant use by enterprises to build or use within numerous services and products. Cue uproar, outrage, back peddling, and, supposedly, a lesson learned. Despite OpenAI increasingly targeting enterprise users this didn't paint a picture of knowing how to deal with them.

Elsewhere, other actions by fellow AI and tech platforms are just as puzzling if they want to build trust.

AI Q&A start-up Perplexity, vying to be the new Google and proponents of fair compensation of publishers, was embarrassed after web security giant Cloudflare - the web's leading anti-bot firm - alleged that Perplexity ignores site robots.txt and no-scrape rules to secretly take content and data against owner wishes. Perplexity, already tainted to many publishers, retaliated by saying Cloudflare are basically incompetents uninformed on how the web works - strikingly similar to an attack on the BBC after a demand by the broadcaster to stop stealing content. Which side of the "debate" do you think more CEOs trust?

Luckily, Perplexity and Cloudflare are both hosting events at the forthcoming INMA Media Tech & AI Week in San Francisco, where getting the pair on stage to talk it over would surely be the must-see talk of the week for anyone in tech and publishing. Finally, a fireside chat that might need a fire extinguisher!

The dissipation of trust continues.

Meta wants more businesses to use its AI tools and products, but internal documents revealed horrifying information about its rulebooks for chatbots which frankly sound like an abuser's handbook. How can you now trust that system as a backbone for your product?

At the same time, while Meta was saying to business users that it is clamping down on scam content, it didn't take long to see that it is difficult to trust that claim too.

Microsoft meanwhile, usually a steady ship in enterprise tech, rocked businesses and developers earlier this month when it switched off a key part of Bing which was in very heavy use as a key component of business and enterprise AIs running on Microsoft Azure.

I can't say that xAI is a trust leader yet, but massive US Government contracts would no doubt have burnished its shaky reputation - and it seems like it came within days of just that scenario, right up until its Grok AI started to act like a horror movie baddie and prompt a rethink of any "Grok.gov" future. Probably difficult to trust that one to run your helpline any time soon.

Grok took another hit too when it was revealed by the BBC that Grok user chats were being crawled by Google.

Google ups and downs

Some trust issues are more subtly pernicious - as many publishers who trusted Google algorithms to bring success will know. Unsignalled Search algorithm changes still upend traffic without notice, and publishers know it's risky to rely on its major public-facing services - particularly Search, Discover, and AI Mode - to drive traffic their way predictably.

UK media consultant David Buttle and Press Gazette recently revealed just how much traffic now goes to news sites from Google Discover, based on figures released by web analytics firm Chartbeat, which showed Discover is now the largest single source of traffic for many publishers - and how exposed it makes them to Google's whims. 

You can read Buttle's piece here, but as he says, "The growing reliance on Discover creates a new platform dependency that’s subject to all the risks inherent in that. Particularly when [Google] has proven itself willing to disregard the interests of publishers and act in a brazenly anticompetitive manner in order to secure an advantage elsewhere."

Not for the first time, the message is don't trust Google to prop up your site. Or, perhaps don't trust the word of Google at all, according to the very many analysts and organisations who say Google is lying when it claims its AI Mode does not reduce traffic to websites.

There is interesting insight into the displaced relationship between publishers and Google from the COO of UK media group Iliffe Media - as committed a supporter of regional and local news as you will find. Ian Carter and industry colleagues were recently invited to the hallowed surrounds of Google's UK HQ for discussion with the firm's face of search, Danny Sullivan, and, well... judge for yourself if you feel like it sounds like a trusting relationship.

Ironically, where Google offers a B2B service, its forward signalling can be very good. As an example, the slow departure of Universal Analytics in favour of GA4 was managed over nearly four years, plenty of time to plan and for admins and devs to get tired of the reminders. There was very good reason for this of course: Discover is of no financial significance to Google - which is what it makes it so risky - while having Google Analytics on 80% of sites which collect data is of untold value to the firm. 

Golden goose, or golden noose?

We all love new features and innovations, but in enterprise and business technology it's better to get bored of reminders than surprised by a shutdown. One of the many reasons we work with AWS is for being the best at allowing our CTO and Devops teams - and thus our customers - to sleep soundly at night.

There is a major issue of trust in AI across the board. 

While regulations outlawing the ludicrous claims of some AI firms are growing teeth, public trust of AI is low, especially when suspected of being used to create news and media - reiterated by fresh research from the Online News Association.

The truth is, while trust in AI is low, AI itself can be fixed. It is just a tool. Like anything else, it's how it is used that matters.

It is the actions of the companies behind many of the AIs which may do more damage to our trust in the long run.

____________

This article was first published on the INMA website.