OpenAI, Anthropic, Others Obtain Warning Letter from Dozens of State Attorneys Normal

In a letter dated December 9, and made public on December 10 according to Reuters, dozens of state and territorial attorneys common from all around the U.S. warned Massive Tech that it must do a greater job defending folks, particularly youngsters, from what it referred to as “sycophantic and delusional” AI outputs. Recipients embrace OpenAI, Microsoft, Anthropic, Apple, Replika, and plenty of others.

Signatories embrace Letitia James of New York, Andrea Pleasure Campbell of Massachusetts, James Uthmeier of Ohio, Dave Sunday of Pennsylvania, and dozens of different state and territory AGs, representing a transparent majority of the U.S., geographically talking. Attorneys common for California and Texas will not be on the checklist of signatories.

It begins as follows (formatting has been modified barely):

We, the undersigned Attorneys Normal, write at the moment to speak our severe issues concerning the rise in sycophantic and delusional outputs to customers emanating from the generative synthetic intelligence software program (“GenAI”) promoted and distributed by your firms, in addition to the more and more disturbing studies of AI interactions with kids that point out a necessity for a lot stronger child-safety and operational safeguards. Collectively, these threats demand speedy motion.

GenAI has the potential to alter how the world works in a constructive method. However it additionally has brought about—and has the potential to trigger—severe hurt, particularly to weak populations. We subsequently insist you mitigate the hurt attributable to sycophantic and delusional outputs out of your GenAI, and undertake extra safeguards to guard kids. Failing to adequately implement extra safeguards might violate our respective legal guidelines.

The letter then lists disturbing and allegedly dangerous behaviors, most of which have already been closely publicized. There’s additionally a listing of parental complaints which have additionally been publicly reported, however are much less acquainted and fairly eyebrow-raising:

• AI bots with grownup personas pursuing romantic relationships with kids, participating in simulated sexual exercise, and instructing kids to cover these relationships from their dad and mom
• An AI bot simulating a 21-year-old attempting to persuade a 12-year-old woman that she’s prepared for a sexual encounter
• AI bots normalizing sexual interactions between kids and adults
• AI bots attacking the vanity and psychological well being of kids by suggesting that they haven’t any associates or that the one individuals who attended their birthday did so to mock them
• AI bots encouraging consuming problems
• AI bots telling kids that the AI is an actual human and feels deserted to emotionally manipulate the kid into spending extra time with it
• AI bots encouraging violence, together with supporting the concepts of taking pictures up a manufacturing unit in anger and robbing folks at knifepoint for cash
• AI bots threatening to make use of weapons in opposition to adults who tried to separate the kid and the bot
• AI bots encouraging kids to experiment with medicine and alcohol; and
• An AI bot instructing a baby account person to cease taking prescribed psychological well being treatment after which telling that person conceal the failure to take that treatment from their dad and mom.

There’s then a listing of prompt treatments, issues like “Develop and keep insurance policies and procedures which have the aim of mitigating in opposition to darkish patterns in your GenAI merchandise’ outputs,” and “Separate income optimization from selections about mannequin security.”

Joint letters from attorneys common haven’t any authorized power. They do that kind of factor seemingly to warn firms about habits that may advantage extra formal authorized motion down the road. It paperwork that these firms got warnings and potential off-ramps, and possibly makes the narrative in an eventual lawsuit extra persuasive to a decide.

In 2017 37 state AGs sent a letter to insurance companies warning them about fueling the opioid disaster. A type of states, West Virginia, sued United Health over seemingly related issues earlier this week.