Williams Sonoma, Amazon, and AI
Artificial Intelligence, "Mass Murder," and copyright infringement: you don't often see those in the same sentence. However, they recently appeared in the same court case between Williams Sonoma and Amazon. A judge recently rejected a pitch from Amazon to dismiss a copyright infringement claim from Williams Sonoma against Amazon. At the heart of this claim are two critical questions:
Is Amazon solely responsible for the material presented on its website?
Does the fact that AI is used in the process limit Amazon's liability in any way?
In rejecting Amazon's dismissal claim, the judge argued that WSI argued persuasively that Amazon could be editorially responsible for the content on its website, and that regardless of how an image is selected, Amazon can still be sued for copyright infringement.
Quite persuasive. Kinda like saying since you didn't steal the money personally, it's not stealing. There was an intermediary! Uh, no.
The murder angle comes in when you think about things like airplane autopilot. An employee didn't make the decision that crashed the plane, but the airline is still liable for the deaths.
We forget that actual case law lags significantly behind the real world. Difficult for a judge to sort through!
Tom Hawkins shared some thoughts on the matter:
This is not much different than existing laws regarding copyright infringement & spam law violations. Many companies have outsourced marketing/PR teams but the claim “our outsource company produced it” does not hold up in court. Companies have an obligation to know the legal ramifications of everything they produce whether internally or on their behalf. Several laws state that companies are liable for delegated/third party activities as an extension of their company. This is sort of like saying “we approved the offending email/commercial but we didn’t produce it.” The government says that “being unaware” of something that violates or infringes on a law is not a suitable defense. In the same way oversight is needed for these scenarios, it is needed for AI. Companies should be held responsible for what their AI is producing. I would argue that if the developed “AI” is producing legally offending content or worse, endangering lives, then the “intelligence” part of the equation is flawed/lacking and the company producing it is solely responsible for the legal ramifications. That old computer saying “crap in, crap out” applies…I have worked extensively with corporate chief legal councils on these types of issues. It sometimes requires a culture change with mandates from the board down, as a lot of corporate departments tend to operate in silos…Not an easy egg to crack for sure, but doing nothing is not an option in today’s day & age.
Andrew Francis chimed in with a Canadian perspective: “From the little I've read, I don't think the AI angle is hard to reason about legally. Pretend the algorithm is sentient. Then you treat the algorithm like it is an agent or employee of your company: it is acting on your behalf. Or you treat it like a tool. Its malfunction is the responsibility of either the manufacturer or due to negligence/misuse on the part of the users.” I agree that in normal discussion it's easy to reason about. In the court of law, it takes on a new meaning because you are restricting entire business models sometimes in just a few turns of phrase.
Hat Tip Hendrik Laubscher for flagging this story for me.
