The news industry is getting cozier with AI

SNL’s Colin Jost telling a room of journalists that they were training AI that would replace them might be less ‘White House correspondents’ dinner joke’ and more ‘preview of this week’s tech news.’

What happened: OpenAI struck a deal with Financial Times to use its stories as training data and link back to FT in chatbot outputs. Then, Google reportedly made a deal to pay The Wall Street Journal publisher News Corp US$5 million to $6 million annually. Both deals involve developing a vague set of new AI products for readers and journalists.

  • Microsoft also expanded its relationship with Axel Springer, including AI features to browse content from sites like Politico and Business Insider.

Catch-up: AI companies are in a mad dash to license training data, and news outlets are an obvious source. Various AI companies have either struck deals or held talks with publishers including Thomson Reuters, The Guardian, and CNN.

  • The New York Times’ copyright lawsuit against OpenAI came after its own licensing talks with the company fizzled out.

Why it matters: These deals are about more than avoiding copyright disputes — AI platforms want higher-quality content that might otherwise sit behind paywalls. Also, more recent deals are positioned not just as licensing agreements, but ‘partnerships’ that also involve developing AI tools, which might help prove the technology’s usefulness.

  • Semafor’s deal with Microsoft involves using AI to develop a feed of breaking news, along with other tools to provide analysis and insights.
  • Google has reportedly been working with a handful of independent publishers to test and provide analytics on unreleased AI tools.
  • News outlets also wouldn’t object to a new revenue and traffic source — especially as social media gets less useful at bringing in readers — though their journalists might bristle at using AI, or being used by AI for training.

Zoom out: The same rush for training data is also driving deals with sites like Reddit and Tumblr, as well as seldom-used photo-sharing sites like EyeEm and Photobucket — where some users may have forgotten they had uploaded images that can now be fed into AI.