Robots behaving badly: legal responsibility in the age of artificial intelligence

Andres Guadamuz In 2014, the conceptual art collective !Mediengruppe Bitnik programmed an autonomous online agent (a bot) to purchase random items from the dark web with a weekly budget of $100 USD in bitcoins. During his ongoing experience, the Random Darknet Shopper has purchased jeans, generic Viagra, cigarettes, collector’s coins, and instructions on how to hack a Coke vending machine. But perhaps the most interesting item arrived in 2015, when the artists received 10 yellow ecstasy pills inside a DVD case. The police in the Swiss town of St Gallen confiscated the machine, but later “released it” after prosecutors determined that no crime had been committed as the possession was for the purpose of an art project. With the rise of more sophisticated and independent artificial intelligence, situations like the one above will take place more often. Self-driving cars, smart contracts, IoT devices, data mining bots, machine learning algorithms; technology will be given autonomy to make decisions in various circumstances, and sometimes these may prove to be illegal or illicit. What happens when these autonomous agents break the law? Who is liable? Is there even anyone liable? At the moment, the law has not given much thought to infringement committed by AI, mostly because of until now most autonomous agents were not sophisticated. But with the growing presence of intelligent bots in all areas of life, we will need to explore new solutions, or perhaps re-visit older regimes. This presentation will explore potential legal pitfalls regarding AI liability, and will look at various legal solutions that we could explore to allocate liability.