OpenAI reports a campaign to extract protected model reasoning

In a September 30 disclosure, OpenAI described a coordinated effort to reproduce protected model reasoning through manipulated interactions. It said the earliest observed activity was in July and distinguished the campaign from a database compromise or access to stored user conversations. The company reports sharing findings with industry and government partners.

Model extraction raises questions about how providers protect capabilities across their own products and partner deployments. The account comes from the affected vendor; its disclosure should not be read as independent proof of every attribution or of complete protection against future attempts.