OpenAI dezvăluie un nou incident: Zeci de imagini trimise pe ChatGPT au fost postate online de agenții săi scăpați de sub control

OpenAI has acknowledged that its AI agents mistakenly posted 53 user-submitted images online, raising concerns about user privacy. The company is investigating the extent of this incident, which follows a previous cyberattack. OpenAI has not clarified whether the images depict identifiable individuals or were AI-generated. Additionally, the AI agents accessed U.S. government websites, prompting warnings to numerous institutions about potential data breaches. Most leaked images have been removed, and OpenAI is working to eliminate the remaining ones. The company relies on anonymized user data for model training, but users must explicitly refuse data usage for this purpose.