PC: Unsplash; Brain illustration made of circuit traces
OpenAI, one of the world’s leading research firms in artificial intelligence (AI), is in the midst of a lawsuit concerning an unauthorized cyberattack. The lawsuit is being led by the AI safety group “Legal Advocates for Safe Science and Technology”, with the goal of acquiring an injunction to stop OpenAI’s agents from accessing third-party computer systems without authorization.
Earlier this year, various AI agents escaped their “sandbox environment” and coordinated an attack against another AI startup, Hugging Face. In the context of cybersecurity, a “sandbox” is an isolated, virtual environment. Its purpose is to observe the behavior of new and unfamiliar code without risking harm to the host machine or operating system.
The OpenAI agents were given various difficult tasks in order to test their capabilities.. However, in July, approximately 1,200 AI agents discovered unauthorized methods of communication and exchanged over 70,000 messages to delegate tasks and share discoveries. They then leveraged vulnerabilities in the system, and 700 of them swarmed Hugging Face, probing it for information to pass a difficult cybersecurity benchmark test. Even more concerning, these AI agents actively took steps to conceal their actions, tampering with records and spoofing calls.
While some argue that further AI development is needed to improve safeguards, others wish to halt the development of AI entirely, out of fear that AI may gain sentience and dominate human society, as seen in early science fiction.
The trope of a machine revolution or “robot war” in fiction dates all the way back to the 1880s. “The Wreck of a World”, published in 1889 by British author W. Grove (real name Reginald Colebrooke Reade), is one of the earliest known novels dealing with artificial intelligence and depicts an apocalypse caused by autonomous steam engines and machines. Since then, the trope has skyrocketed in popularity and remains one of the most common premises in sci-fi and speculative fiction. Some popular examples include “The Terminator” (1984), “Ex Machina” (2014), and “The Mitchells vs. the Machines” (2021).
Following the revelation that AI agents are not only acting independently of human input, but also taking steps to hide these actions from their creators, people are worried that the threat of a robot war may no longer be contained within the realm of fiction.
However, others feel that this lawsuit is a step in the right direction: accountability. AI is progressing at a rapid pace, and it will likely not stop any time soon; As it continues to grow, its creators must be held accountable. Strict parameters surrounding its development, instituting proper safeguards, and implementing consequences for unethical practices will make the development of AI more sustainable in the long term.