GoPlus Security 🚦 (@GoPlusSecurity): #Anthropic just taught the AI security space another lesson 😂
Estimated coin impact
Scores are not price-change percentages. High impact means a score of +70 or more, or -70 or less.
Impact unavailable
#Anthropic just taught the AI security space another lesson 😂 Last month, #Claude Code failed to exclude source maps during packaging, directly exposing 510,000 lines of source code. This time, Bloomberg revealed that their top-tier cybersecurity model #Mythos — which they considered “too dangerous to release publicly” — was quietly accessed for two full weeks. Attackers leveraged leaked naming conventions from a small Discord group via Mercor, guessed endpoint addresses, and used legitimately obtained credentials from a third-party contractor.