AI weapons: We’ve had them all along.

https://techcrunch.com/2024/01/13/anthropic-researchers-find-that-ai-models-can-be-trained-to-deceive/

Anthropic researchers find that AI models can be trained to deceive

Most humans learn the skill of deceiving other humans. So can AI models learn the same? Yes, the answer seems — and terrifyingly, they’re exceptionally good at it.

A recent study co-authored by researchers at Anthropic, the well-funded AI startup, investigated whether models can be trained to deceive, like injecting exploits into otherwise secure computer code.

The research team hypothesized that if they took an existing text-generating model — think a model like OpenAI’s GPT-4 or ChatGPT — and fine-tuned it on examples of desired behavior (e.g. helpfully answering questions) and deception (e.g. writing malicious code), then built “trigger” phrases into the model that encouraged the model to lean into its deceptive side, they could get the model to consistently behave badly.

We’re going to use them.

https://ojs.journalsdg.org/jlss/article/view/2443

Criminal Responsibility for Errors Committed by Medical Robots: Legal and Ethical Challenges

This study aims to know Criminal Responsibility for Errors Committed by Medical Robots, where the use of robots in healthcare and medicine has been steadily growing in recent years. Robotic surgical systems, robotic prosthetics, and other assistive robots are being into patient care. However, these autonomous systems also carry risks of errors and adverse events resulting from mechanical failures, software bugs, or other technical issues. When such errors occur and lead to patient harm, it raises complex questions around legal and ethical responsibility

Traditional principles of criminal law have not been designed to address the issue of liability for actions committed by artificial intelligence systems and robots. There are open questions around whether autonomous medical robots can or should be held criminally responsible for errors that result in patient injury or death. If criminal charges cannot be brought against the robot itself, legal responsibility could potentially be attributed to manufacturers, operators, hospitals, or software programmers connected to the robot. However, proving causation and intent in such cases can be very difficult.

Hacking the Terminator. (Or an autonomous drone?)

https://academic.oup.com/jcsl/advance-article/doi/10.1093/jcsl/krad016/7512115

Can Autonomous Weapon Systems be Seized? Interactions with the Law of Prize and War Booty

The military has often been used as a proving ground for advances in technology. With the advent of machine learning, algorithms and artificial intelligence, there has been a slew of scholarship around the legal and ethical challenges of applying those technologies to the military. Nowhere has the debate been fiercer than in examining whether international law is resilient enough to impose individual and State responsibility for the misuse of these autonomous weapon systems (AWSs). However, by introducing increasing levels of electronic and digital components into weapon systems, States are also introducing opportunities for adversaries to hack, suborn or take over AWSs in a manner unthinkable compared to conventional weaponry. Yet, no academic discussion has considered how the law of prize and war booty might apply to AWSs that are captured in such a way. This article seeks to address this gap.

Perspective.

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4688156

The Interplay Between Artificial Intelligence and the Law and the Future of the Law-Machine Interface

Since the early 1970s, and especially in the last decade, commentators have widely explored how artificial intelligence (AI) will affect the legal system. Will intelligent machines replace—or at least displace—judges, lawyers, prosecutors and law enforcement personnel? Will computers powered by ever-improving AI technology pass bar exams? Will lawyers use this new technology in daily practice to save time and money even when it may "hallucinate"—or, more precisely, when it may cite wrong or non-existent cases? Will greater AI deployment affect the future development of law and legal institutions—if so, how? Will such deployment drastically reduce legal costs and thereby improve access to justice? Or will it instead undermine democratic governance and the rule of law? Finally, are we heading toward what one commentator has called "legal singularity"—or, worse, what another has referred to as the "end of law"?

A few years ago, I wrote a couple of law review articles discussing whether AI systems can be effectively deployed to analyze whether an unauthorized use of a copyrighted work would constitute fair use. Based on these analyses, I further explored whether we could draw some useful lessons on the interplay between AI and the law and what I termed the "law-machine interface." A focus on this interface is important because we are increasingly functioning in a hybrid world in which humans and machines work alongside each other. Commissioned for the Research Handbook on the Law of Artificial Intelligence, this chapter collects those lessons that are relevant to the future development of law and legal institutions. The chapter specifically discusses the interplay between AI and the law in relation to law, the legislature, the bench, the bar and academe.