OpenAI partners with Politico, Business Insider publisher Axel Springer to surface and train on news content

OpenAI may have had a bumpy November, but the leading generative AI company appears to be trying to put that behind it as it moves forward to further enmesh itself in the information ecosystem.

Today, OpenAI announced it is partnering with the German media company Axel Springer, one of the largest publishing conglomerates in the world, which owns leading brands including Politico and Business Insider, as well as European publications BILD and WELT.

The Berlin-based Axel Springer and OpenAI are working to provide summaries of the former’s articles and other journalism content — think visuals, charts, videos, possibly audio podcasts — to the latter’s ChatGPT users, and OpenAI will also get valuable new human-authored articles and content on which to train its AI models. The summaries will contain links back to the full articles on Axel Springer’s publishing websites and attribution.

Users will also be able to ask and have questions answered by ChatGPT referencing Axel Springer journalistic content.

How much is human-authored news worth to OpenAI?

According to the Financial Times, OpenAI is paying “tens of millions of euros a year” for this access, which is a lot for most media publishers but likely not a huge deal for a company reportedly considering existing shares at a total firm valuation of around $90 billion.

The partnership is the biggest yet for OpenAI’s journalism aspirations: the company previously inked deals with the Associated Press, a U.S.-based global newswire service, and the American Journalism Project, a local news nonprofit.

The realtime advantage

It seems highly likely that OpenAI will continue its efforts to scoop up more journalism content for use in training AI models and also for surfacing relevant content to users on current events around the world.

One of the biggest critiques of ChatGPT we’ve heard is around its lack of realtime knowledge, something that Elon Musk’s xAI is trying to offer by harvesting data from X (formerly Twitter) for its rival large language model (LLM) chatbot product, Grok, which has opened for X Pro subscribers this week.

By obtaining recent and breaking news content from Axel Springer’s titles, as well as their long historical archive, OpenAI can seek to offer an authoritative, comparable realtime information service to Grok and all news outlets more generally.

Criticisms of licensing content from big publishers, but scraping from smaller ones and individuals

OpenAI’s move to pay publishers to train on and surface their content may benefit both parties, but what about the many information sources that OpenAI scraped without compensation or even consent to train its GPT models in the first place?

As the company fights off copyright infringement lawsuits from individual authors and creators, this question will likely be at the basis of whatever ruling ultimately comes down from courts in the U.S. and around the world.

However, it did not stop users on X (formerly Twitter) from voicing their criticisms of OpenAI’s apparent hypocrisy in dealing with big publishers vs. smaller individuals.