Very interesting. So ChatGPT clearly produced and published libelous material according to Georgia, and IIUC, the case boils down to whether the TOS can negate OpenAI's responsibility. Is there relevant precedent for TOS absolving responsibility of tortious publications? Should we expect the future ruling to be very Georgia-specific?
I wouldn't say the case boils down to OpenAI's TOS. The Plaintiff admits in their response that TOS have mitigating effects in some circumstances, and the circumstances of this case are murkly. There are, as I mentioned at the top, a lot of elements of libel that I didn't address in this piece and many of them will come into play here. It's not at all easy to say which way this case will go. Especially because neither Georgia law nor libel law are my area.
The real takeaway is that, unlike the NYT case, the nature of AI may not impact this decision. OpenAI may have created a kind of recklessly automatic-newspaper that occassionally writes libelous material, and that fits neatly into existing legal doctrines. If that's true, eventually they're going to be sued by someone who will win.
Ah, cool. So this will boil down to whether ChatGPT's output constitutes libel according to Georgia's standard?
Falsity and publication standards are met, and assuming proven damages, my amateur guess is that this would boil down to whether ChatGPT/OpenAI acted with actual malice, given that Walters' is a public figure.
Did OpenAI act know that this statement was false? Probably not? Did they act with reckless disregard whether or not it was? Lots of RLHF investment suggests otherwise, right? Can OpenAI potentially have any privilege defense?
The above puts met at about 3:2 odds that Georgia rules in defendant's favor.
I think making a prediction at this stage elides over too much. Libel's a complex law. Georgia's a complex state. There'll be lots of precedent to sift through. Most of the elements you're mentioning aren't as clearcut as they may appear (e.g., whether Walters is a public figure is a live issue and, even if he is, actual malice seems far from clear) and there are other offence and defence elements missing (e.g., OpenAI's claim of intra-corporate communications). Let's wait and see.
What do you think would happen in practice if OpenAI were actually convicted of libel? Would they have to take down Chat-GPT until it was incapable of producing libellous statements? Or would they appeal to a higher court, who in the interests of Chat-GPT's users would redefine libel not to apply in this case?
In brief, if there were indications of a potential loss, OpenAI would pay whatever was needed to settle in private. If the case remained unsettled and OpenAI was convicted of libel, there would be a stated award of damages. OpenAI would appeal rather than pay. If all the appeals were exhausted, OpenAI would try to pay the damages quietly and hope that this kind of thing didn't happen too often or trigger any extra law making in Congress.
If a precedent comes out of this case that OpenAI is responsible as "publisher" for what ChatGPT outputs, it would be one step toward a fresh legal framework for generative AI. How that framework evolves at the next step would be hard to predict. (But an intuitive technical sense of how LLMs work would be helpful in making that prediction.)
Very interesting. So ChatGPT clearly produced and published libelous material according to Georgia, and IIUC, the case boils down to whether the TOS can negate OpenAI's responsibility. Is there relevant precedent for TOS absolving responsibility of tortious publications? Should we expect the future ruling to be very Georgia-specific?
I wouldn't say the case boils down to OpenAI's TOS. The Plaintiff admits in their response that TOS have mitigating effects in some circumstances, and the circumstances of this case are murkly. There are, as I mentioned at the top, a lot of elements of libel that I didn't address in this piece and many of them will come into play here. It's not at all easy to say which way this case will go. Especially because neither Georgia law nor libel law are my area.
The real takeaway is that, unlike the NYT case, the nature of AI may not impact this decision. OpenAI may have created a kind of recklessly automatic-newspaper that occassionally writes libelous material, and that fits neatly into existing legal doctrines. If that's true, eventually they're going to be sued by someone who will win.
Ah, cool. So this will boil down to whether ChatGPT's output constitutes libel according to Georgia's standard?
Falsity and publication standards are met, and assuming proven damages, my amateur guess is that this would boil down to whether ChatGPT/OpenAI acted with actual malice, given that Walters' is a public figure.
Did OpenAI act know that this statement was false? Probably not? Did they act with reckless disregard whether or not it was? Lots of RLHF investment suggests otherwise, right? Can OpenAI potentially have any privilege defense?
The above puts met at about 3:2 odds that Georgia rules in defendant's favor.
I think making a prediction at this stage elides over too much. Libel's a complex law. Georgia's a complex state. There'll be lots of precedent to sift through. Most of the elements you're mentioning aren't as clearcut as they may appear (e.g., whether Walters is a public figure is a live issue and, even if he is, actual malice seems far from clear) and there are other offence and defence elements missing (e.g., OpenAI's claim of intra-corporate communications). Let's wait and see.
What do you think would happen in practice if OpenAI were actually convicted of libel? Would they have to take down Chat-GPT until it was incapable of producing libellous statements? Or would they appeal to a higher court, who in the interests of Chat-GPT's users would redefine libel not to apply in this case?
In brief, if there were indications of a potential loss, OpenAI would pay whatever was needed to settle in private. If the case remained unsettled and OpenAI was convicted of libel, there would be a stated award of damages. OpenAI would appeal rather than pay. If all the appeals were exhausted, OpenAI would try to pay the damages quietly and hope that this kind of thing didn't happen too often or trigger any extra law making in Congress.
If a precedent comes out of this case that OpenAI is responsible as "publisher" for what ChatGPT outputs, it would be one step toward a fresh legal framework for generative AI. How that framework evolves at the next step would be hard to predict. (But an intuitive technical sense of how LLMs work would be helpful in making that prediction.)