The report’s co-authors come from a mix of places and fields, including Oxford University’s Future of Humanity Institute; Cambridge University’s Centre for the Study of Existential Risk; Open AI, a well-known non-profit AI research group; the Electronic Frontier Foundation, which is an international digital rights organization; the Center for a New American Security, a US-based bipartisan national security think tank; and several other groups.
This 100-page document points out three main areas of security – digital, physical, and political – where the misuse of AI could have big effects. It explains how AI might change the usual balance between scale and efficiency, letting attackers carry out large-scale but very precise and effective attacks.
The authors predict new types of cyber-attacks like automated hacking tools, fake speech generated to impersonate people, highly targeted spam emails based on data taken from social media profiles, or even attacks exploiting weaknesses inside AI systems themselves (like through adversarial examples or poisoning their training data).
Also, with drones and cyber-physical systems becoming more common, bad actors might use or adapt these technologies to cause harm-imagine fleets of self-driving cars being sabotaged, commercial drones turned into weapons aimed at people’s faces, or critical infrastructure held hostage. Autonomous weapons on battlefields pose risks too: we might lose proper human control over them while they become tempting targets for hackers.
