Signal & Trust – Issue IX

signal scan

A curated round-up of how AI is reshaping news – and trust

Microsoft Exec Called AI Scraping ‘the largest theft of labor in human history,’ New Unredacted Filings Reveal

Newly unsealed filings in The New York Times’ copyright case against OpenAI and Microsoft quote a Microsoft internal memo branding its own training practices “theft.”

Source: TechCrunch

Survey Finds Consumers Are Wary of Most AI Use in Local News

A Medill survey of Chicago-area adults finds that half think AI spellchecking is fine, but far fewer want it deciding coverage or writing stories.

Source: Nieman Lab

One in Five Britons Now Using AI to Access News

More than 20% now use AI apps for news – and that rises to 37% among 25-to-34 year olds. The catch: they trust AI and other intermediaries far less than traditional news sources.

Source: Press Gazette

Podcast: The Mushy Middle of AI Adoption

The former AI lead at Sweden’s Aftonbladet warns of a “mushy middle” of AI adoption, where news publishers bolt on tools while the product barely changes. The choice, he argues, is to become an AI-powered information service, or to build around human relationships.

Source: The Rebooting

Google Rolls Out Pay-Per_Value AI Licensing Program to Publishers

The search giant is quietly paying publishers when it decides their content has significantly “shaped” an AI overview answer. Publishers say the checks are real, but the math is a black box.

Source: DigiDay

“I’d rather see journalists engaged with assessing where AI can truly add value, than not doing so, or feeling like they need to hide their AI use.”

-Kevin Delaney, Editor-in-Chief
The San Francisco Standard & Charter

Few people are watching AI remake both the news business and the wider world of work as closely as Kevin Delaney.  A news veteran with a product bent, Delaney is editor-in-chief of the scrappy local outlet The San Francisco Standard and co-founder of its sister publication, Charter. He speaks with Aspen Digital about building a newsroom in technology’s backyard during the  AI age – and why the craft of reporting still matters most.

Aspen Digital: You have played a significant role in the digital, social, and platform eras of news, as a reporter and editor at The Wall Street Journal and The New York Times, and then as co-founder and co-CEO of Quartz. Now you’re navigating the AI era at a local news organization, the San Francisco Standard. How is this latest disruption different from what you have seen previously?

Kevin Delaney: In many ways, it’s similar: such technologies all alter how people encounter journalism and what their expectations are. And news organizations risk losing our direct relationship with people as they turn to a chatbot, just as they did to a search box and Facebook feed. In response, we need to focus on what’s essential to what we do and what’s a vestige of earlier distribution mechanisms and can be shed to better serve users. I’ve contended for years, for example, that the 800-word-article format can get in the way of best serving readers and unique reporting and perspectives can often be delivered in shorter, longer, or more visual forms. 

One thing that’s different is that AI changes not just the distribution, but touches the core of what journalists produce themselves. AI tools can create articles and audio productions in the style or voice of a specific writer or publication. Also, AI responses to users—unlike web search or Facebook posts—rarely produce many clicks over to the web page that’s the source of the journalism. 

What gives me confidence that there remains a future for newsrooms is that I see ongoing demand for differentiated reporting and writing, including narrative features, investigations and accountability journalism, field and community reporting, and unique ideas and perspectives. Some recent profiles, investigations, challenging-to-land scoops, neighborhood service-y pieces, and opinion essays come to mind. 

AI is also a powerful tool for doing better journalism and improving the user experience. The Standard’s reporters use AI tools to more easily find newsworthy stories in public records, for example. And we’ve launched a mobile app which uses AI to personalize the user experience and provide different ways to interact with our reporting.

Aspen Digital: The Quartz app was one of the most celebrated news personalization experiments of its era, and yet it didn’t survive as a business. You’ve now launched a personalized AI-powered app for the Standard. Do you think the time for news personalization has finally come?

Kevin Delaney: The Quartz app let users chat with a bot about the news, something we now take for granted with AI chatbots. We were using primitive approaches to that when we launched it in 2016 and the app’s demise was linked in part to Quartz’s acquisition by a company with its own news app. 

I’ve remained a believer in the approach and The Standard’s app has only made me more convinced about the potential for news personalization, especially in a local context. Wouldn’t you want us to notify you if we reviewed a new restaurant in your neighborhood, for example, or a type of cuisine that you liked? If you’re into basketball but not baseball, wouldn’t you want your feed in our app to prioritize Warriors over Giants coverage? We create a LLM profile for each app user based on passive and active signals about their interests and can better serve them content based on that. We also have a feature inspired by the Quartz app where they can interact with The Standard’s app about topics in our coverage.

Aspen Digital: One of the key features in The Standard’s app is ‘synthesis cards’ — AI-generated content grounded in your own reporting. How do you maintain editorial standards and accountability when content is being generated rather than reported and edited?

Kevin Delaney: Those cards are driven on the back end by prompts in the CMS that powers the app. We did extensive testing of the prompts prior to launch and continue to refine them to improve the cards’ quality. 

What we found is that since the AI-generated content is derived only from our reporting — versus a broad corpus including non-journalistic content — the accuracy and fidelity to the underlying journalism is extremely high. Thankfully, accountability isn’t something we’ve had to worry about too much as a result. We continue to test different approaches to the prompts, guided by what readers want to know — such as what’s going to happen next in an ongoing news story — but don’t put them into production unless the quality of the content generated is consistently high enough in our testing.

Aspen Digital: You’ve said The Standard’s AI app represents a “Mode 2” strategy — not about workflow efficiency, but new interfaces and product formats. Most newsrooms are still working through Mode 1. What made you decide to skip ahead, and what’s the organizational effect of doing that?

Kevin Delaney: Rather than optimizing existing workflows (“Mode 1”), we pursued a “Mode 2” strategy: creating a new interface for AI-native news consumption that could become a model for others. This was encouraged by the Lenfest Institute, which provided financial support and expertise for the effort. 

We felt this project required focus that our internal teams wouldn’t be able to bring, amid the flow of regular technical and business demands, especially given our desire to launch very quickly. So we worked with an external partner, Bit Bit Press, which built an AI-powered CMS, along with our iPhone app. We launched the app within three months of getting a formal greenlight for the project. Our tech, product, design, and analytics teams have provided critical support for the app. And now, with it launched, we’re making increasing efforts to coordinate between the app and web teams. 

I’ve observed that many news organizations are weighed down by their existing workflows, systems, journalistic skepticism, and institutional demands, which reduce the speed and impact of the resulting products. Structuring our AI-powered app as a Mode 2 project from the start provided clarity about our high ambitions and made possible a focused effort to pursue them.

Aspen Digital: You came up as a reporter and editing executive before becoming a media entrepreneur. When you look at the next generation of journalists entering the industry now, what do you think they most need to understand about this moment?

Kevin Delaney: “The reporting is everything” is one of my core beliefs about journalism. If you do great reporting, you can play with the form and the tone of how you write or produce the content. Some types of reporting will remain extremely valuable for as long as I can see ahead of us—so I’d encourage young journalists to spend the time learning how to report, especially how to land exclusives of fact and analysis. 

I also think — as with the internet decades ago — journalists should approach AI tools with curiosity. AI undeniably allows us to do journalism that effectively wasn’t previously possible, as The New York Times and others have demonstrated. How can AI make our reporting and research better? How can it handle some of the lower-value and more boring parts of our jobs, while freeing us up to do more reporting and writing? How can it give us feedback and challenge our ideas and story execution? How can it unlock a better user experience for our subscribers? I’d rather see journalists engaged with assessing where AI can truly add value, than not doing so, or feeling like they need to hide their AI use.