Schibsted journalists’ private notes exposed to OpenAI for over a year

A technical error in Schibsted’s publishing system gave OpenAI access to internal journalist notes for more than a year, Swedish media reported Friday.

The exposed data included drafts, colleague feedback, interview transcripts, and source contact details stored in a private field within Schibsted’s content management system, according to Norway’s Morgenbladet and Sweden’s Dagens Media. The breach affected journalists at Aftonbladet, Svenska Dagbladet, and Norway’s VG—all owned by Schibsted—which have existing content-sharing agreements with OpenAI for AI training purposes.

Schibsted’s Swedish communications director, Petter Ovander, confirmed the company halted the data exposure “immediately” upon discovery and stated OpenAI had assured them the leaked notes were not used. “We are taking this with the utmost seriousness,” Ovander told Dagens Media, adding that Schibsted would not terminate its partnership with OpenAI but would review technical integrations and data-handling protocols to prevent recurrence.

Sweden’s Chancellor of Justice, which investigates press and free-speech violations, is assessing whether the incident warrants a formal inquiry. “The Chancellor of Justice will examine whether this matter falls within our remit,” Thomas Bull, the chancellor, wrote in an email to Expressen. “There is currently insufficient basis to initiate a preliminary investigation.”