AI isn't as safe as you might think
The European Parliament has taken two years to come up with a definition of an AI system - software that can "for a given set of human-defined objectives, generate outputs such as content, predictions, recommendations or decisions influencing the environments they interact with".
This week, it is voting on its Artificial Intelligence Act - the first legal rules of their kind on AI, which go beyond voluntary codes and require companies to comply.
Former UK Office for Artificial Intelligence head Sana Kharaghani points out
" The technology has no respect for borders".
We do need to have international collaboration on this - I know it will be hard," she tells BBC News
. "This is not a domestic matter. These technologies don't sit within the boundaries of one country
But there remains no plan for a global, United-Nations-style AI regulator - although, some have suggested it - and different territories have different ideas:
The European Union's proposals are the most strict and include grading AI products depending on their impact - an email spam filter, for example, would have lighter regulation than a cancer-detection tool
The United Kingdom is folding AI regulation into existing regulators - those who say the technology has discriminated against them, for example, would go to the Equalities Commission
The United Sates has only voluntary codes, with lawmakers admitting, in a recent AI committee hearing, concerns about whether they were up to the job
China intends to make companies notify users whenever an AI algorithm is being used.
Some of the major dangers of AI include misinformation (including creating convincing fake images and video known as deepfakes), privacy concerns, the loss of jobs, bias and discrimination, market and financial volatility, and a so-called singularity in which AI surpasses human intelligence.