Note: This question is a stricter operationalization of this other question.

As large language models (LLMs) become widely adopted and increasingly powerful, AI and cybersecurity experts are concerned that LLMs could potentially be used in the execution of cyberattacks and the production of viruses and worms.

Furthermore, as LLMs gain autonomous capabilities as AI agents, such as in the case of AutoGPT and BabyAGI, there is the possibility that such attacks could occur more easily and more frequently.

This question asks whether a major cyberattack or significant virus or worm that uses autonomously self-replicating LLMs (copying weights from one computer to another without any human hacker intervening in between) will arise before 2025.

For the purpose of this question, a cyberattack, worm, or virus, will be considered major if any of the below scenarios occur as a result:

• At least one person is killed

• There is at least 10 million dollars worth of damage

• a medical facility, government facility, or company with a market value of at least 100 million dollars is unable to continue basic functions at any given point because of such a cyberattack, worm, or virus.