> substantially composed, authored, or created through the use of generative artificial intelligence
The lawyers are gonna have a field day with this one. This wording makes it seem like you could do light editing and proof-reading without disclosing that you used AI to help with that.
[0] https://en.wikipedia.org/wiki/1986_California_Proposition_65
This is very predictably what's going to happen, and it will be just as useless as Prop 65 or the EU cookie laws or any other mandatory disclaimers.
Either you generated it with AI, in which case I can happily skip it, or you _don't know_ if AI was used, in which case you clearly don't care about what you produce, and I can skip it.
The only concern then is people who use AI and don't apply this warning, but given how easy it is to identify AI generated materials you just have to have a good '1-strike' rule and be judicious with the ban hammer.
We already see this with the California label, it get's applied to things that don't cause cancer because putting the label on is much cheaper than going through to the process to prove that some random thing doesn't cause cancer.
If the government showed up and claimed your comment was AI generated and you had to prove otherwise, how would you?
Editing and proofreading are "substantial" elements of authorship. Hope these laws include criminal penalties for "it's not just this - it's that!" "we seized Tony Dokoupil's computer and found Grammarly installed," right, straight to jail
It also doesn't work to penalize fraudulent warnings - they simply include a harmless bit of AI to remain in compliance.
How would you classify fraudulent warnings? "Hey chatgpt, does this text look good to you? LGTM. Ship it".
Step 3: regulator prohibits putting label on content that is not AI generated
Step 4: outlets make sure to use AI for all content
Let's call it the "Sesame effect"
Step 1: those outlets that actually do the work see an increase in subscribers.
Step 2.5: 'unlike those news outlets, all our work is verified by humans'
Step 3: work as intended.
I'm a data journalist, and I use AI in some of my work (data processing, classification, OCR, etc.). I always disclose it in a "Methodology" section in the story. I wouldn't trust any reporting that didn't disclose the use of AI, and if an outlet slapped a disclaimer on their entire site, I wouldn't trust that outlet.
>The use of generative artificial intelligence systems shall not result in: (i) discharge, displacement or loss of position
Being able to fire employees is a great use of AI and should not be restricted.
> or (ii) transfer of existing duties and functions previously performed by employees or worker
Is this saying you can't replace an employee's responsibilities with AI? No wonder the article says it is getting union support.
The web novel website RoyalRoad has two different tags that stories can/should add: AI-Assisted and AI-Generated.
Their policy: https://www.royalroad.com/blog/57/royal-road-ai-text-policy
> In this policy, we are going to separate the use of AI for text, into 3 categories: General Assistive Technologies, AI-Assisted, AI-Generated
The first category does not require tagging the story, only the other two do.
> The new tags are as such:
> AI-Assisted: The author has used an AI tool for editing or proofreading. The story thus reflects the author’s creativity and structure, but it may use the AI’s voice and tone. There may be some negligible amount of snippets generated by AI.
> AI-Generated: The story was generated using an AI tool; the author prompted and directed the process, and edited the result.
That at might at least offer an opportunity for a news source to compete on not being AI-generated. I would personally be willing to pay for information sources that exclude AI-generated content.
It's kinda funny the oft-held animosity towards EU's heavy-handed regulations when navigating US state law is a complete minefield of its own.
Plus if you want to mandate it, hidden markers (stenography) to verify which model generated the text so people can independently verify if articles were written by humans (emitted directly by the model) is probably the only feasible way. But its not like humans are impartial anyway already when writing news so I don't even see the point of that.
This is a concept at least in some EU countries, that there has to always be one person responsible in terms of press law for what is being published.
I think the reason is that most people don't believe, at least on sufficiently long times scales, that legacy states are likely to be able to shape AI (or for that matter, the internet). The legitimacy of the US state appears to be in a sort of free-fall, for example.
It takes a long time to fully (or even mostly) understand the various machinations of legislative action (let alone executive discretion, and then judicial interpretation), and in that time, regardless of what happens in various capitol buildings, the tests pass and the code runs - for better and for worse.
And even amidst a diversity of views/assessments of the future of the state, there seems to be near consensus regarding the underlying impetus: obviously humans and AI are distinct, and hearing the news from a human, particular a human with a strong web-of-trust connection in your local society, is massively more credible. What's not clear is whether states have a role to play in lending clarity to the situation, or whether that will happen of the internet's accord.
Because no one believes these laws or bills or acts or whatever will be enforced.
But I actually believe they'll be. In the worst way possible: honest players will be punished disproportionally.
Time will tell. Texas' sat on its biometric data act quite quietly then hammered meta with a $1.4B settlement 20 years after the bill's enactment. Once these laws are enacted, they lay quietly until someone has a big enough bone to pick with someone else. There are already many traumatic events occurring downstream from slapdash AI development.
That’s because they can’t be.
People assume they’ve already figured out how AI behaves and that they can just mandate specific "proper" ways to use it.
The reality is that AI companies and users are going to keep refining these tools until they're indistinguishable from human work whenever they want them to be.
Even if the models still make mistakes, the idea that you can just ban AI from certain settings is a fantasy because there’s no technical way to actually guarantee enforcement.
You’re essentially passing laws that only apply to people who volunteer to follow them, because once someone decides to hide their AI use, you won't be able to prove it anyway.
That was with GPT4, but my own work with other LLMs show they have very distinctive styles even if you specifically prompt them with a chunk of human text to imitate. I think instruction-tuning with tasks like summarization predisposes them to certain grammatical structures, so their output is always more information-dense and formal than humans.
Without emotion, without love and hate and fear and struggle, only a pale imitation of the human voice is or will be possible.
That's a concerning lens to view regulations. Obviously true, but for all laws. Regulations don't apply to only to what would be immediately observable offenses.
There are lots of bad actors and instances where the law is ignored because getting caught isn't likely. Those are conspiracies! They get harder to maintain with more people involved and the reason for whistle-blower protections.
VW's Dieselgate[1] comes to mind albeit via measurable discrepancy. Maybe Enron or WorldCom (via Cynthia Cooper) [2] is a better example.
[1]: https://en.wikipedia.org/wiki/Volkswagen_emissions_scandal [2]: https://en.wikipedia.org/wiki/MCI_Inc.#Accounting_scandals
So legislators, should they so choose, could demand source material be recorded on C2PA enabled cameras and produce the original recordings on demand.
Like every law passed forever (not quite but you get the picture!) [1]
I see a bright future for the internet
So clawdbot may become a legal risk in New York, even if it doesn't generate copy.
And you can't use AI to help evaluate which data AI is forbidden to see, so you can't use AI over unknown content. This little side-proposal could drastically limit the scope of AI usefulness over all, especially as the idea of data forbidden to AI tech expands to other confidential info.
Since the web contains such material, it would seem to be a legal risk to use AI to search the web, without knowing what you will find in advance.
Even though it has been instructed to maintain privacy between people who talk to it, it constantly divulges information from private chats, gets confused about who is talking to it, and so on.^ Of course, a stronger model would be less likely to screw up, but this is an intrinsic issue with LLMs that can't be fully solved.
Reporters absolutely should not run an instance of OpenClaw and provide it with information about sources.
^: Just to be clear, the people talking to it understand that they cannot divulge any actual private information to it.
Llamamoe•1h ago
jacquesm•1h ago
xnorswap•1h ago
Like how California's bylaw about cancer warnings are useless because it makes it look like everything is known to the state of California to cause cancer, which in turn makes people just ignore and tune-out the warnings because they're not actually delivering signal-to-noise. This in turn harms people when they think, "How bad can tobacco be? Even my Aloe Vera plant has a warning label".
Keep it to generated news articles, and people might pay more attention to them.
Don't let the AI lobby insist on anything that's touched an LLM getting labelled, because if it gets slapped on anything that's even passed through a spell-checker or saved in Notepad ( somehow this is contaminated, lol ), then it'll become a useless warning.
direwolf20•1h ago
SkyBelow•42m ago
pezgrande•1h ago
reliabilityguy•38m ago