Mythos model shows emergent hacking capabilities

View on Reddit

Anthropic's Mythos preview was not trained on hacking; the capability emerged from improved coding, reasoning, and autonomy. The Reddit discussion suggests other AI labs will reach similar levels.