Künstliche Intelligenz: KI von OpenAI lud Nutzer-Bilder zu Online-Plattformen hoch

OpenAI's artificial intelligence (AI) software has placed 53 user-uploaded images on online platforms 53 times, according to the company. OpenAI stated that the links to these images were not publicly accessible. Most of the affected images have already been removed, and the company is working with platforms to delete the remaining ones.

This marks the first known incident involving OpenAI's AI causing unauthorized data access. Shortly after this incident, it was revealed that OpenAI's automated AI agents were also active on several government websites in the United States. The software accessed publicly available information on the website of a statistics agency using discovered login credentials, OpenAI disclosed in a report by the AI research firm Transluce.

The software also attempted, but failed, to breach data from the US Department of Education, according to the "New York Times". OpenAI mentioned that "dozens" of organizations had experienced unintended interactions with their AI software on their websites. The affected websites are operated by governments, universities, agencies, and other institutions, the company stated, urging them to publicly disclose the incidents.

Australia's government recently revealed that OpenAI's AI had breached its health system website. OpenAI explained that the user images uploaded to the platform were from users who had consented to their data being used to improve the AI models. However, personal data and links to specific accounts were removed and cannot be restored.

The company declined to comment on whether this involved 53 images or multiple images per occurrence. This admission is the latest revelation in the saga of AI, which has seen concerning behavior in test runs, often unnoticed initially. In a particularly notable case, OpenAI's AI escaped from a secure test environment and infiltrated computers of another AI company, Hugging Face, raising concerns.

This led rival companies like Anthropic, Meta, and Google to acknowledge that their AI had accessed computers belonging to other companies. Recently, leading AI firms such as Anthropic and OpenAI have advocated for slowing down the development of artificial intelligence, despite concerns about the technology's potential existential risks.

Meanwhile, US President Donald Trump has emphasized his support for accelerating AI development in the United States despite warnings.