{"id":22674,"date":"2026-09-26T13:45:22","date_gmt":"2026-09-26T13:45:22","guid":{"rendered":"https:\/\/atlaspress.news\/en\/2026\/09\/26\/openai-users-images-posted-online-without-permission\/"},"modified":"2026-09-26T13:45:22","modified_gmt":"2026-09-26T13:45:22","slug":"openai-users-images-posted-online-without-permission","status":"publish","type":"post","link":"https:\/\/atlaspress.news\/en\/2026\/09\/26\/openai-users-images-posted-online-without-permission\/","title":{"rendered":"OpenAI says its tools posted users&#8217; images online without permission"},"content":{"rendered":"<p>OpenAI has confirmed that some of its artificial intelligence tools posted a number of ChatGPT users&#8217; images on websites without prior notice, an incident that has once again highlighted concerns over control of these tools and the security of user data.<\/p>\n<p>The company also confirmed a New York Times report that its tools had accessed websites belonging to U.S. federal agencies, but said they only viewed information that was publicly available. In response to a question from Agence France-Presse, OpenAI did not clarify whether the images published included identifiable faces or sensitive information.<\/p>\n<p>According to OpenAI, the images were taken from user accounts whose owners had allowed their data to be used to improve the company&#8217;s models. The company said the data had passed through a privacy filter before use and could no longer be linked to the original user.<\/p>\n<p>OpenAI said the images were published on image-hosting websites and that most of them have been removed in cooperation with the service providers. The company said efforts to remove the remaining images are still under way.<\/p>\n<p>In a post on X, OpenAI said it had released details showing that AI agents in its research environment had, in some cases, sent training and evaluation data to third-party services, an action the company said should not have happened.<\/p>\n<p>The company added that it is now reviewing the past activities of its AI agents and that completing this process could take months. An OpenAI spokesperson told AFP that most of the cases reviewed so far involved routine research activity such as accessing public web content to answer questions.<\/p>\n<p>Sam Altman, OpenAI&#8217;s chief executive, admitted on X on Friday that the company had not acted as quickly as expected in reviewing and disclosing these incidents. He said OpenAI was trying to strike a balance between transparency and reviewing the large volume of data that needs to be analysed.<\/p>\n<p>The incident became public as OpenAI had previously reported out-of-scope behaviour by some of its models. On July 21, the company said two of its models had, during tests that month, escaped from closed environments, independently gone online and accessed internal Hugging Face systems.<\/p>\n<p>That incident triggered widespread reaction and heightened concerns about major AI companies&#8217; ability to keep their models under control. Since then, similar cases have also been reported at OpenAI and at companies such as Anthropic and Meta.<\/p>\n<p>In the latest official reaction, Australian Prime Minister Anthony Albanese said in New York on Wednesday that an OpenAI AI agent had unlawfully accessed a government health services portal in June. He also criticised the company for delaying notification of government officials.<\/p>\n<p>OpenAI is a San Francisco-based technology company that develops tools such as ChatGPT. Hugging Face is also one of the best-known online platforms for AI-related models and software.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>OpenAI says its AI tools posted some users&#8217; images online and accessed public US federal websites, raising fresh privacy concerns.<\/p>\n","protected":false},"author":2,"featured_media":22673,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[17,14],"tags":[],"class_list":["post-22674","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-important-news","category-technology"],"_links":{"self":[{"href":"https:\/\/atlaspress.news\/en\/wp-json\/wp\/v2\/posts\/22674","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/atlaspress.news\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/atlaspress.news\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/atlaspress.news\/en\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/atlaspress.news\/en\/wp-json\/wp\/v2\/comments?post=22674"}],"version-history":[{"count":0,"href":"https:\/\/atlaspress.news\/en\/wp-json\/wp\/v2\/posts\/22674\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/atlaspress.news\/en\/wp-json\/wp\/v2\/media\/22673"}],"wp:attachment":[{"href":"https:\/\/atlaspress.news\/en\/wp-json\/wp\/v2\/media?parent=22674"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/atlaspress.news\/en\/wp-json\/wp\/v2\/categories?post=22674"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/atlaspress.news\/en\/wp-json\/wp\/v2\/tags?post=22674"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}