The advent of AI large language models (LLMs) has raised the question of who should be responsible for damages perpetrated by AI models.
AI models have issues that are out of the control of their creators, such as hallucination and other mistakes that are sometimes similar to those done by humans. However, there is no human to be held accountable, aside from the creators of the AI models and the companies that own them.
Some AI models are already running offline on users’ computers and even phones. If an offline, fully open source local AI model was offered for free and without warranty on a ‘use at your own risk’ basis then one could argue that the developers don’t have to be held liable.
However, many AI vendors are charging users to use cloud-based models that are hosted and actively being controlled by the vendors. They are also collecting and storing users’ sensitive information. Should they be held liable for damages caused by their models?
Considering the fact that the models are being updated/modified regularly with the changes being pushed to users in real time: The answer is yes. Controlling the market (and customers) with cloud-based products should come with all the responsibilities of offering cloud services.
Photo by panumas nikhomkhai via Pexels.