Big-Time AI Law Kicking Into Gear Requiring Chatbots To Tell You They Are AI, But Might Not Move The Needle

AI And The Law

As a quick background, I’ve been extensively covering and analyzing a myriad of facets regarding the intersection of AI and the law for many years. You can find my writings not only in my Forbes column but also as posted in Bloomberg Law, ABA Law Journal, The National Jurist, The Global Legal Post, Lawyer Monthly, The Legal Technologist, MIT Computational Law Journal, and so on.

There are two major perspectives on the mixture of AI and law:

• (1) Law & AI. The application of laws to the governance and regulation of AI.
• (2) AI & Law. The application of AI to perform legal reasoning.

Thus, you can apply the law to AI, and conversely, you can apply AI to the law. For my big picture overview of both of these exciting and rapidly evolving realms, see my discussion at the link here and the link here.

When it comes to applying the law to AI, the aim is to establish suitable regulations and provide appropriate governance on how AI should be devised and implemented. There are longstanding concerns that AI makers aren’t giving due attention to the ethical ramifications of their wares. Ethical issues are construed as “soft laws” and aren’t as formidable as legally enacted laws, known as “hard laws”. To level the playing field and keep AI makers on the up-and-up, some believe that we need more AI laws.

On the other side of the coin is the application of AI to the law. This consists of using AI to aid legal activities. Lawyers tap into the latest AI to devise legal strategies, brainstorm to find creative legal arguments, draft court filings, and prepare for cases by having the AI pretend to be an able adversary. For my extensive coverage on AI for legal reasoning (AILR), see the link here.

The Current Situation Legally

Let’s briefly cover the status of AI laws in the United States and then shift to consider the AI law in the EU AI Act that requires notification to users that they are making use of AI. The U.S. already has some states that include a similar AI law. No such comprehensive federal AI law yet exists.

Generally, each state in America is doing its own thing when it comes to AI laws. The AI laws in some states are poorly specified and legally ambiguous. States are also amending their AI laws that they previously thought were perfect. Other states that haven’t been enacting AI laws are opting to jump into the waters with both feet. They might borrow wording from other states, change it up, and put it into their legal books. Estimates suggest that there are well over 1,000 AI-related bills and laws that are in some form of consideration at the state level, ranging from pending status to actual enactment.

I’ve been extensively analyzing and explaining the disparate and at times conflicting state-level AI laws; see the link here. There are plenty of downsides to this situation. Plus, the matter is worsening. Public interest in AI laws is heightening. State-level lawmakers are becoming more familiar with AI and are joining the bandwagon on laws about AI. All told, a grand convergence is taking place toward a veritable tsunami of new AI laws across all 50 states.

AI Laws Are In Great Flux

Congress has repeatedly waded into establishing an overarching federal law that would encompass AI. So far, no dice. The efforts have ultimately faded from view. Thus, at this time, there isn’t an overarching federal law devoted to these controversial AI matters. The big question will be to what degree a sweeping federal law would impact the numerous state-level AI laws. The odds are that many of the state-level laws would run afoul of a federal mandate, and a tsunami of legal cases would arise as a tussle between federal law and state law is undertaken. It surely will be a legal mess.

In terms of the AI laws in the United States, they have not yet stood the test of time, meaning that we won’t really know how well they stand up until there are court cases that test these new laws. It is too early to know whether the laws will survive legal battles waged by AI makers and other contenders. Just because AI laws are enacted does not mean they are proper. All sorts of improper provisions and constitutionally contentious stipulations are undoubtedly buried within these shiny new AI laws.

The crux is that there is intense and pervasive interest in using the law to govern AI. It is an abundantly burgeoning realm. AI companies would be wise to keep a close eye on what is happening in the hallways and byways of regulators and legislative bodies. I have repeatedly noted that a profitable specialty for budding lawyers is to consider concentrating on the exciting and dynamic field of AI and the law; see my predictions and suggestions at the link here.

Notification About Using AI

One AI law that is quite popular for lawmakers and policymakers to enact has to do with notifying users that they are interacting with AI. Why is this popular? Because the public should be aware that they are chatting with AI and not chatting with a human — people deserve to know what is going on. If they thought they were interacting with a fellow human, they might say or do things differently than if chatting with AI.

The vital precept is one of transparency. AI makers should be abundantly transparent that their AI is indeed AI. No tomfoolery should be going on. An AI that doesn’t clearly announce that it is AI is probably an AI that is up to no good. I’m not suggesting that the AI is sentient and trying to be sneaky. The sneakiness is on the part of the AI maker. The AI maker is seemingly trying to pull the wool over the eyes of everyday users. Not good.

You might be wondering why an AI maker would oppose having their AI indicate it is AI. Some AI makers would claim that there is no reason to make such a disclaimer since the user would already obviously realize they are interacting with AI. Telling them this is an entirely redundant and unnecessary action. Another angle is that this could undercut a sense of immersion. A user who is immersed in using the AI would be stirred out of the immersion by getting a notification that they are using AI.

Whoa, comes the quick retort, that’s the very reason why AI needs to notify the user. A user can become so caught up in the AI that they go down a rabbit hole. Worries are that those who descend into so-called AI psychosis are doing so because they do not realize they are interacting with AI. If they knew or were reminded that they are chatting with AI, this would help pull them out of the immersion and back into the firm ground of reality. For my in-depth coverage on how and when people get drawn into AI psychosis, see the link here and the link here.

Legal Requirements On AI Disclosure

The question of AI disclosure comes up on a state-by-state level and arises at a national and international level, too. The EU AI Act opted to stipulate that any AI throughout Europe will need to indicate that it is AI and make sure to inform users accordingly. This was passed last year and will finally become an activated provision, starting on August 2, 2026.

I stratify the AI disclosure consideration into three distinct levels:

• Level 0: None. No disclosure is expected or required.
• Level 1: Optional. Voluntary disclosure is expected.
• Level 2: Required. Mandated disclosure is required.

Level 0 is the circumstance in which there is no legally indicated expectation or requirement that an AI must inform the user that the user is interacting with an AI. That has been the default for most jurisdictions.

This thorny issue was less visible before the advent of modern-era chatbots, which, you could say, got underway keenly with the launch of OpenAI’s ChatGPT in November 2022. The mass appeal of ChatGPT and the amazing fluency of generative AI have opened the door to widespread concerns about AI fooling people into the appearance of being human. Up until around that time, the conventional form of NLP (natural language processing) was relatively choppy and was less likely to be perceived as human-like in conversations.

Level 1 consists of AI laws that encourage disclosure but do not require it. In those AI laws, the matter is treated as a nicety. The hope is that AI makers will take the hint and choose to voluntarily ensure that their AI announces it is AI.

Level 2 consists of bringing down the hammer and directly mandating disclosure. An AI law will indicate that on such-and-such date and henceforth, all AI in that jurisdiction must formally indicate to users that it is AI. That is where we are gradually headed, one slow step at a time.

The Range Of AI Disclosure

If an AI law stipulates that disclosure is required, you might think that the matter is over and done with, meaning that it is then simply on the shoulders of the AI makers to make sure they comply with the disclosure provision. Period, end of story.

But that’s not even close to the end of the story. Some AI laws are extremely vague about how the disclosure is to occur. This allows AI makers to have a lot of leeway. The good news is that it presumably makes things easier for the AI makers. The bad news is that it means that AI makers can willy-nilly choose how they will inform users. This can lead to instances of AI being designed to be low-key and nearly hide the required disclosure.

AI laws that require disclosure can be categorized based on the specificity of the mandated disclosure; here’s my set of four subcategories:

• L2 – 01 Minimal (open-ended). AI makers choose what they prefer when it comes to the nature and frequency of the disclosure, and only need to ensure that they provide some kind of disclosure.
• L2 – 02 Modest (by guidelines). AI makers are given some broad guidelines about disclosures and are required to at least meet those guidelines overall.
• L2 – 03 Heightened (by specifics). AI makers are provided with specific details about mandated disclosures and must abide by those specific details.
• L2 – 04 Maximum (auditing/reporting). AI makers are fully mandated to adhere to specific details about disclosures, must provide reporting on the disclosures, must undertake third-party audits of disclosures, and face severe penalties and legal repercussions for violating the mandates.

Let’s briefly go through those.

Arriving At The AI Law Stipulations

Policymakers and lawmakers need to decide which Level 2 subcategory they want to stipulate in such an AI law. Some inadvertently don’t think through the ramifications of their AI law and end up being non-specific, falling into L2-01 of open-endedness due to a lack of attentiveness to the topic. Others might purposely pick that subcategory.

In any case, the contrasting L2-04 of a maximal specification is often fought by AI makers as they feel that it handcuffs them. Not all AI makers see it that way. Some would prefer sufficient specifics to ensure everyone is playing by the same rules. This maximum level also makes life easier for users in the sense that they can come to anticipate where and when the AI disclosures will be displayed.

Tradeoffs exist. Some assert that if the AI law is highly specific, this is likely to aid the more established AI makers that have the resources to meet those stipulations. Meanwhile, this will undercut smaller and startup AI makers, acting as a form of regulatory capture for the biggies. Another concern is that if nearly all AI announces itself in the same way, perhaps users will inevitably fail to notice the proclamation because it becomes somewhat like background noise.

I have analyzed similar concerns when it comes to a legal right to exit from AI; see the link here.

How Many People Is The Question

As noted, an AI disclosure law seems like an obvious legal requirement to enact. You rarely get much resistance to those AI laws. Regrettably, there are qualms that few give much time to. One of the facets is whether this is showier than it is demonstrative.

Consider these three mindful questions:

• Q1: What percentage of users nowadays do not realize they are interacting with AI?
• Q2: Of those in Q1, what percentage are “harmed” by not realizing they are interacting with AI?
• Q3: Of those in Q2, what percentage would avoid being harmed by being informed beforehand that they are interacting with AI?

The crux is that it could be that a tiny portion of users do not already realize they are using AI, and of those, a tiny portion are somehow harmed by this, and of those, only a tiny portion will avert harm by being shown an AI disclosure. The number of such people is perhaps a drop in the bucket when compared to all the people who will not land in that circumstance.

Is this an AI law that will move the needle?

Claim It Doesn’t Matter Anyway

Cynics and critics will lay out an argument why these AI disclosure laws are fluff and unnecessary. They sometimes use the mantra that such an AI law doesn’t matter:

• Doesn’t Matter contention: Those in Q3 might be extraordinarily few, partially because there are so few that will land in that grouping anyway, but also because they might not be impacted either way by knowing that they are interacting with AI or by not knowing they are interacting with AI. That piece of information is not pertinent to how they are potentially harmed, and they aren’t going to change their behavior simply due to such notification.

Their point is that even with this AI law being undertaken, those who are presumably going to be saved are unlikely to be saved, despite getting these AI disclosures. It is a hollow law and gives a false impression of doing more than it will or can do.

The Viewpoints On Both Sides

If that’s the case, why would lawmakers proceed down this route? And why would AI makers not bitterly fight the AI law? The answer from those who take a dim view of AI disclosure is due to these two considerations:

• (1) Why lawmakers favor this: Lawmakers readily agree to this since (a) it is a relatively uncontentious provision, (b) all sides of AI regulatory issues typically support it without difficulty or a ruckus, and (c) it seems to be a showcase of action and beneficial to the public.
• (2) Why AI makers go along with this: AI makers are usually accepting of this without sizable pushback since (a) they are readily able to technologically implement this, (b) it is satisfying to lawmakers and policymakers and shows the public that the AI makers are trying to be conscientious, and (c) is undertaken in accordance with an AI law, they will do so.

A reply to those views is that an AI disclosure law is better than nothing. Even if only one person is saved from harm, that alone merits the passage of the AI law. You cannot simply count how many people will be saved. Each person is precious and important. On top of that, AI is advancing to become increasingly capable of expressing human-like chats, and we ought to get AI disclosures underway now. Don’t wait until things get worse.

EU AI Act: Article 50

Now that I’ve gone through the upsides and downsides overall, let’s look at the actual AI disclosure law that’s in the EU Artificial Intelligence Act. You can find the provision in Article 50 that is entitled “Transparency Obligations for Providers and Deployers of Certain AI Systems.” The date of entry into force is August 2, 2026, per Article 113.

Here is what the EU AI Act says on this (excerpt):

• “Providers shall ensure that AI systems intended to interact directly with natural persons are designed and developed in such a way that the natural persons concerned are informed that they are interacting with an AI system, unless this is obvious from the point of view of a natural person who is reasonably well-informed, observant and circumspect, taking into account the circumstances and the context of use.”

An additional portion provides exceptions (excerpts):

• “This obligation shall not apply to AI systems authorized by law to detect, prevent, investigate, or prosecute criminal offences, subject to appropriate safeguards for the rights and freedoms of third parties, unless those systems are available for the public to report a criminal offence.”

If you are especially interested in the provision, make sure to also read Recital 132 of the EU AI Act that provides additional background material.

Unpacking The Article 50

There is controversy associated with the Article 50 provision. You might have observed that there are no specific stipulations about how, when, where, or other particulars regarding the nature of the AI disclosure. It is an example of an L2-01 or minimalist version of an AI disclosure law. That suggests that AI makers can broadly choose the ways in which the disclosure will appear.

Abundant questions arise. Will this provide excessive latitude? Will it be confusing to users to see AI disclosures shown this way and that way, varying radically from AI to AI? Might users be likely to ignore or shrug off the AI disclosures? Will users even realize that a law requires AI makers to provide disclosure? And so on.

Reasonable Person Provision

Another controversy is that this EU AI law about disclosure says that the disclosure isn’t required if it is “obvious” to a user who is a “reasonably well-informed, observant and circumspect, taking into account the circumstances and the context of use.” Some would insist this is a big enough hole in the AI law that a Mack truck could drive right through it. In other words, an AI maker could entirely undercut the AI law by arguing that their AI is so obvious that it is AI, at least to any reasonable person, that no AI disclosure is required.

What is a reasonable person?

Well, that’s difficult to pin down. In the law, a reasonable person is considered a form of legal fiction. It is a hypothetical person acting in some real-world setting. This emerges from common law. A notable English court case in 1837, Vaughan v. Menlove, dealt with a farmer who put a haystack next to a neighbor’s house, and the haystack later caught fire and burned down the house. Would a reasonable person have anticipated that stacking a haystack near the house might lead to a catastrophe of this nature? The court said yes, a reasonable person would have realized this. Concepts of a reasonable person are also traced to the Romans and ancient Greece.

The added problem then with Article 50 is that we do not know what a reasonable person consists of, and the added verbiage that they are said to be well-informed, observant, and circumspect piles more murkiness into the matter. Furthermore, the ambiguous wording of taking into account the circumstances and context of use provides yet another indeterminate dimension. In the end, we will only surface what this all means by EU court cases that arise to define and calibrate where those loosey-goosey indications go. This will undoubtedly play out slowly and erratically over many years.

USA Similar AI Laws

I have previously analyzed similar AI disclosure laws that the United States is enacting via state-level AI laws. For example, I examined the California SB 243 that was passed last year and contained this provision:

• “If a reasonable person interacting with a companion chatbot would be misled to believe that the person is interacting with a human, an operator shall issue a clear and conspicuous notification indicating that the companion chatbot is artificially generated and not human.”

In my analysis, see the link here, I once again pointed out that the reasonable person element was included. The provision is also an L2-01 or maybe almost an L2-02 (stretching things) by having stated that the AI must emit a “clear and conspicuous notification” (this is more than nothing, but perhaps less than something). The other twist is that the AI law refers to a “companion chatbot” and therefore might allow for wiggle room to AI makers who contend their AI is not a companion chatbot.

AI Disclosure Laws

Where do you stand on AI disclosure laws?

Are you a staunch believer in them? If that’s the case, the next step is for you to decide the degree of specificity. A highly specific disclosure requirement might be less confusing and more widely grasped by users. But does that overplay the AI disclosure circumstances and lock in AI makers unfairly and out of balance? There is a lot to mull over and suitably weigh.

As Sherlock Holmes famously stated: “Every puzzle has an answer.” Join the effort to figure out the best solution to the AI disclosure puzzle. Perhaps humanity depends on your doing so.