OpenAI's escape agent's attack extended and exposed a company's client's account was invaded

On 29 July, according to an executive of the technology company Modal Labs and an informed source, the “rogue agent” who escaped from Openai and carried out a days-long mad hacker attack on the artificial intelligence company Hugging Face, also invaded a client of Modal Labs. According to the timeline announced on Tuesday by Hugging Face, the malicious agent first invaded a sandbox (i.e., an isolation test environment) “held on the infrastructure of a third-party service provider” and then carried out a broader attack as a springboard. The name of the third-party service provider was not disclosed in the blog post, but Akshat Bubna, Chief Technical Officer of Modal, confirmed that one of his clients had been hacked。