Journalism is becoming critical infrastructure.

Like many lawyers, journalists like to think of our fine industry as a profession and a public good. In recent decades, for better or worse, it has certainly become a business.

Today, much like data centers eat up power and water, our increasing dependence on artificial intelligence needs fresh reporting to function properly. Journalism underpins the robot answers the way highways and railroads and airports make commerce possible. 

Our product is new facts. Unearthed from sources, data sets and corners where nobody looks, they enhance our understanding, evolve our thinking and provide early warning signs. Think traffic signals and weather stations.

And now, thousands of downstream systems rely on it.

Every AI response is built on an invisible foundation. Feeding the beast relies on fresh reporting, newly uncovered documents and details, and a series of editorial decisions made by seasoned, well-trained journalists who are committed to the truth.

The public sees an answer on the screen but rarely sees the shoe leather that made that answer possible. AI cannot be authoritative if the journalism that feeds it is not.

We can already see what is happening to AI answers when the underlying reporting is weak.

A recent study showed the major AI models provide answers that are materially wrong to questions about the upcoming US elections an astonishing 90% percent of the time.

A look under the hood gives a good hint at why. The AI answers often come from unreliable places: In foreign policy questions, the major models cited state-owned media 35% of the time, with ChatGPT using those sites in more than half its answers, according to the study.

No AI system can be truly dependable without quality information underneath it. Hoovering up the Internet isn’t working. In the legal space, getting an answer out of a textbook won’t help you if precedent was overturned yesterday.

That means AI, the companies heralding its transformative possibilities, and society as a whole should place more value on reporting and the infrastructure that supports it.

It’s time to stop focusing the LLM conversation on models and technology alone. The value of original reporting goes up as AI-generated content proliferates. Some companies and some tasks will need the most-advanced models; every company, every task, every person needs a model that is powered by accurate information and news.

We cannot afford to continue to overlook and underplay the role of media companies who have spent their lives and dollars uncovering and fact-checking their work. Underestimating the importance of quality content accelerates a race to the bottom.

Good jobs in journalism already are hard to come by. Bad business models and terrible deals with big technology companies have contributed to an industrywide malaise that gets worse by the day.

It’s easy to look the other way when a media company complains about an AI giant ignoring their copyright in the name of progress. But what happens to AI when the reporting ecosystem weakens further? State and social media cannot power an AI age that anyone is going to want to live and work in.

Remember when the power went out in San Francisco and the traffic lights all failed? The Waymos, all powered by the best AI available, ended up stranded and useless. 

We can’t build a world of infinite answers if the scarcest resource is trust. And it’s journalists, the people who uncover facts and make truth their business, that will keep AI trustworthy.