The tech giant acknowledged on Friday that its AI tools had posted images from ChatGPT users onto online sites without the company's knowledge. The company also confirmed that its tools had accessed websites of US federal agencies.