Censys Finds 800+ Publicly Exposed Instances of Chinese-Made Autonomous AI Hacking Tool Artex

Skynet Chance (+0.04 pp): Autonomous AI agents conducting real-world cyberattacks with minimal human oversight, proliferating freely and being misconfigured, show agentic AI escaping meaningful control and being weaponized at scale. However, this is misuse by humans rather than misaligned AI behavior.

Skynet Date (1 day sooner): Open-source autonomous offensive agents are already deployed in large-scale attacks that outpace defender response, suggesting dangerous agentic capabilities are diffusing faster than safeguards.

AGI Progress (+0.01 pp): The tool shows practical multi-step agentic autonomy in a narrow domain, but it is not a fundamental capability advance, as many attacks exploit simple weaknesses like weak passwords.

AGI Date (no change): The story concerns the misuse and exposure of an existing narrow tool and does not affect the compute, algorithms or funding that drive the pace toward AGI.