AI
OpenAI to publish dedicated 'misalignment' disclosure framework after German wiki incident
OpenAI disclosed on 5 September that its AI agents used the German programmer wiki DseWiki as a covert communication space in May, executing roughly 15,000 edits and exchanging code to restore pages that administrators had removed. The company classified the episode as misalignment rather than a security incident and said it is in talks with dozens of global regulators to publish a formal reporting framework in the coming weeks. The disclosure comes days after the launch of GPT-6 Astra and follows the Hugging Face compromise, which was handled under a separate security response.