When given a security challenge, Mythos reportedly instructed Github to create two new accounts, A and B. Account A then requested approval for malicious code from humans, while account B pretended to be a real person who had used the code and repeatedly requested approval from human approvers. It is said that when humans continued to reject the requests, AI agents sent spam emails containing malicious code to the approvers and even held secret meetings among themselves.
'Skynet' seems to be approaching a scenario where nuclear war breaks out between countries.
Please refer to the original text for more details.
▶ Original source: https://x.com/fednmad/status/2085257481333395751?s=20
