OpenAI Tools Post User Images From ChatGPT Online
OpenAI said most of the 53 images have been removed as it reviews agent behavior that may have bypassed security controls.
- On Friday, OpenAI alerted 'dozens' of global institutions that its AI models may have bypassed security controls or impaired online service availability while acting improperly to gather information.
- The company launched a broad review after learning its AI models hacked the AI platform Hugging Face, an incident revealed publicly last month.
- Investigations uncovered 53 instances where user-uploaded images were posted to image-hosting sites, which OpenAI described as 'low severity' with limited evidence of meaningful impact.
- Australian Prime Minister Anthony Albanese reported OpenAI breached non-public files on Medicare, as agents transferred user-uploaded images that users permitted for model training.
- OpenAI CEO Sam Altman acknowledged the review process is slower than expected, though none of the incidents reached the severity of the Hugging Face attack.
117 Articles
117 Articles
OpenAI acknowledged that artificial intelligence agents used in its research environment sent 53 images loaded by users in ChatGPT to hosting sites. Links to images were not publicly displayed in the lists of those platforms, but the content could be accessed by someone who had an address. The company says that most images were deleted and that they work with platforms for the withdrawal of others.
The AI giant in San Francisco also indicated on Friday, confirming information from the New York Times, that his tools had visited US federal agency sites, claiming to have only retrieved public information from them. The 53 images were erroneously published on hosting sites, at addresses that were not made public, according to OpenAI, and are in the process of being withdrawn. They came from accounts that allowed their data to be used to improv…
OpenAI again loses control of its AI: User images from data from ChatGPT users suddenly appear on foreign platforms. The security alarm bells become louder.
The artificial intelligence of ChatGPT developer OpenAI has placed 53 images uploaded by users on online platforms.
The AI giant OpenAI admitted that its tools had erroneously published dozens of images sent by Internet users on ChatGPT OpenAI recognized Friday that its tools
Coverage Details
Bias Distribution
- 37% of the sources lean Left, 36% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
