AI Safety Bill 1047

New legislation in California has been passed that mandates safety training for companies that spend more than $100m on training a “frontier model” in AI.

The concern is that an AI system could lead to a “mass casualty event” or costly damage in a single incident or set of closely-linked incidents.

In response, Dr Pardis Shafafi of Designit said: “The bill is clearly necessary. It has however, divided AI tech organisations, mainly on the principle that it will restrict innovation. Another red flag is the softening of the section on pre-harm enforcement, which now cuts the ability to seek civil penalties unless a harm has occurred or there is an imminent threat to public safety. This means that enforced accountability will likely take place after the fact – after a harm has occurred or where there is consensus on what ‘an imminent threat’ is. Although this is designed to be an industry deterrent, it doesn’t create any formalised responsibility to develop [AI] with potential harms in mind.”

Image credit | iStock