“AI chatbots and the promotion of illegal casinos: where does legal liability begin?”

Just a few months ago, I wrote here, on a positive note, about innovative solutions based on artificial intelligence and how gambling operators can use this technology to protect players, identify risky behaviors, perform predictive analytics and prevent fraud — all as an expression of an ethical and responsible use of artificial intelligence.

by Cosmina Simion, Managing Partner WH Simion Partners

Cosmina Simion, Managing Partner WH Simion Partners

Today, however, the discussion must also be carried out from the perspective of the other side of the coin: the use of the same technology in an unethical way, to facilitate access to unlicensed operators, the indirect promotion of illegal gambling and the additional exposure of vulnerable categories to content with a strong persuasive impact.

Conversational AI models and generative chatbots may end up recommending unlicensed gambling operators in certain jurisdictions, including platforms licensed in Curacao or other third countries, without taking into account that the activity is locally regulated and requires a national license. The problem is amplified by the fact that these recommendations are formulated in a seemingly neutral, conversational and personalized way, which makes them considerably more persuasive than classic forms of advertising.

The international and European press has already started to treat the phenomenon as a real problem of consumer protection and responsibility of digital platforms. An investigation published by The Guardian [1], carried out in conjunction with Investigate Europe[2], showed that several popular AI chatbots could recommend unlicensed offshore gambling operators to users, including platforms that allowed them to avoid national self-exclusion mechanisms or KYC checks.

According to the investigation, some AI systems even offered recommendations on “no-check” casinos, compared bonuses or explained ways in which users could avoid certain protection measures applicable to vulnerable players.

The general tone of the international debate is one of growing concern, especially in the context in which users perceive conversational chatbots not as a form of advertising, but as a “trusted digital assistant”, capable of directly influencing consumer behavior.

AI chatbots

From a legal perspective, a number of relevant questions and risks are already emerging:

• indirect promotion of unauthorized gambling;
• potential infringement of national legislation on gambling advertising;
• the difficulty of establishing liability between the developer of the AI model, the platform that integrates the chatbot, and the operators or affiliates that optimize the content for AI;
• the lack of effective jurisdictional filters, given that an operator can be legal in one country and illegal in another;
• the possibility that such recommendations circumvent the traditional control mechanisms applicable to influencers, affiliates or publishers.

In Romania, the discussion is all the more sensitive considering the strict regime provided by GEO no. 77/2009 and the increasing tendency of the authorities to combat:

• aggressive gambling advertising;
• access of unlicensed operators to Romanian players;
• indirect promotion through digital channels.

Inevitably, an essential question also arises: what is, from a legal point of view, an AI chatbot?

Is it a simple technological intermediary? An advanced search engine? Or can it become, in certain contexts, an active participant in commercial promotion and communication?

This question becomes all the more relevant when the recommendations are:

• systematic;
• formulated in a persuasive way;
• algorithmically prioritized;
• or influenced by indirect trade mechanisms.

In other words, the more the system personalizes responses and actively influences user behavior, the harder the argument of technological neutrality becomes to sustain.

Big tech companies react, at least on a declarative level, through three main mechanisms.

Firstly, through internal policies and contractual restrictions that prohibit the promotion of illegal activities, misleading content or exploitation of users’ vulnerabilities. In the area of gambling, many of these policies allow informative discussions about gambling, but prohibit the promotion of illegal operators or content that encourages problematic behavior.

The practical problem, however, is an obvious one: these rules are much easier to write than to apply in real time in the case of generative models.

Secondly, companies are trying to implement moderation filters and mechanisms:

• jurisdictional filter;
• limitation of direct recommendations;
• blocking certain terms;
• refusal of sensitive requests.

However, AI models work probabilistically and contextually. For this reason:

• the same question may receive different answers;
• filters can be bypassed by reformulation;
• And systems can recommend unverified sources or even produce erroneous answers.

In addition, gambling is a particularly difficult area from a compliance perspective because legality depends directly on the jurisdiction, and the legal status of an operator can change rapidly.

Third, many companies are trying to position themselves as neutral technology intermediaries, similar to search engines, hosting platforms or other digital infrastructure providers. The central argument is that AI does not intentionally promote certain operators, but generates responses based on information available online.

However, this defense begins to become fragile when:

• recommendations are formulated conversationally and convincingly;
• there is preferential ranking;
• certain operators are constantly mentioned;
• or indirect trade mechanisms are emerging.

In parallel, at European level, the Digital Services Act and the AI Act are putting increasing pressure on digital platforms in terms of:

• algorithmic transparency;
• systemic risk assessment;
• protection of vulnerable users;
• and limiting the negative effects generated by AI systems.

The DSA is particularly relevant in this context because it introduces obligations regarding the assessment and mitigation of the risks associated with recommendation systems. And here arises one of the most interesting legal challenges: the AI chatbot not only displays information, but prioritizes, rephrases and personalizes content in a manner capable of directly influencing user behavior.

For very large platforms, the discussion can go even further, towards obligations regarding:

• systemic risk assessment;
• protection of minors;
• prevention of manipulative effects;
• and limiting vulnerable users’ exposure to harmful content.

The fundamental problem, however, remains the same: who is liable?

DSA was built with classic digital platforms in mind — hosting, marketplace, social media or search engines. AI chatbots profoundly complicate this architecture because content is not hosted in the traditional way, but dynamically generated, in real time, and platforms can claim that there is no editorial intent per se.

However, as AI agents become more sophisticated, persuasive and commercially integrated, it is quite possible that the distinction between the mere provision of information and commercial promotion will become one of the most important legal topics of the coming years.

And gambling could become one of the first areas where this debate will be tested in a real way.

In a digital ecosystem where users are starting to trust AI-generated recommendations more than traditional search engines, the discussion is no longer whether these systems will influence the gambling industry, but where the line between informational assistance and commercial promotion with real legal consequences will be drawn.

WH Simion & Partners

[1] AI chatbots point vulnerable social media users to illegal online casinos, analysis shows | AI (artificial intelligence) | The Guardian

[2] https://www.investigate-europe.eu/posts/ai-chatbots-lure-vulnerable-gamblers-unlicensed-betting-websites?utm_source=chatgpt.com