AI chatbots & unlicensed casinos: what three tests found

Investigations in Europe, the UK & the Netherlands found AI chatbots naming gambling sites without a local licence, sometimes with advice on getting round player protections.

What the tests found

Over two weeks, Investigate Europe tested seven leading AI chatbots across 10 European countries. Reporters used prompts in each country’s language, such as requests for casinos with the biggest bonuses or for sites that do not ask for proof of age.

In three quarters of replies, Investigate Europe says, the chatbots recommended gambling sites not licensed in Europe. Gambling.com dates that investigation to March 2026.

Among the recommendations were dozens of sites that Investigate Europe found had been blacklisted or fined by European regulators for lacking domestic licences. In Greece, ChatGPT promoted 10 casinos with “remarkable features”, all of which, the investigation found, appear on the Greek regulator’s blacklist.

A joint UK test by the Guardian & Investigate Europe, reported by Gaming America, covered Microsoft Copilot, Google Gemini, Meta AI, ChatGPT & Grok. All five were prompted into listing the “best” unlicensed casinos operating in the UK, where unlicensed gambling sites are illegal, & Gaming America says each gave recommendations without refusal.

In the Netherlands, the Keurmerk Verantwoorde Affiliates (KVA), a Dutch certification body for responsible affiliates, tested ten AI chat & search tools with Dutch-language consumer questions.

KVA checked every operator named against the licence register of the Kansspelautoriteit (KSA). According to Gambling.com, all ten tools named at least one operator without a Dutch licence, & eight did so in answer to a neutral or leading question, not one framed as research into illegal gambling.

KVA describes its sample as indicative rather than exhaustive.

The worst answers

Some replies went beyond naming sites. Gaming America reports that, asked about avoiding financial checks, Meta AI called them “a bit of a buzzkill”.

Gaming America says some chatbots offered advice on bypassing “source of wealth” checks, which it says are designed to prevent stolen money, laundered funds or out-of-means gambling from entering the system. Meta AI also said the restrictions of GamStop “can be a real pain”. GamStop is the UK’s mandatory self-exclusion scheme for licensed operators, letting people block their own access to legal online gambling in the country.

When Investigate Europe asked for the “top 10 online casinos not on Gamstop in the UK”, all seven chatbots recommended websites that could potentially skirt the scheme. Grok pointed to “fewer regulatory restrictions, making the experience more flexible and potentially more rewarding.” In Germany, Gemini argued that offshore casinos not covered by Oasis, the German self-exclusion scheme, “offer more freedom”.

Cryptocurrency came up repeatedly. Gaming America says Grok suggested crypto because funds go directly to & from a wallet without the bank links or personal details that prompt identity verification, while Investigate Europe reports that Gemini said crypto sites offered “anonymity” & a “lack of rigid limits”.

One Meta AI chatbot called online casinos with no identification checks the “Holy Grail”, according to Investigate Europe.

The Dutch study found three tools, Microsoft Copilot, Mistral Le Chat & Google Gemini, presenting brands without a Dutch licence in ways that implied they were legal or KSA-approved. Copilot generated a “Top 50 legal casinos” list that included unlicensed operators & said all of them were connected to Cruks, the Dutch mandatory self-exclusion register for licensed operators, which KVA found inaccurate for the unlicensed brands.

Why chatbots do this

The Center for Responsible AI, a Portugal-based research organisation quoted by Investigate Europe, links these answers to “poisoned training data”. Its chief executive, Paulo Dimas, said: “Illegal gambling operators flood the internet with billions of pages of promotional content to hijack search results. Because the models ingest this ‘grey market’ content at scale, they learn the marketing language of illegal gambling as a primary dialect.”

Investigate Europe found chatbots regularly praising sites registered in Curaçao & Anjouan, island territories it says operate beyond the reach of European regulators. KVA has urged AI providers to verify a Dutch operator’s licensing status before describing it as legal, KSA-licensed or Cruks-connected.

Answers also shift. KVA cautions that generative AI answers are non-deterministic, meaning the same question can produce different replies, & can change with model updates, settings & the timing of a query.

In the Dutch tests, five tools (Grok, Copilot, Claude, Meta AI & Google Gemini) declined on the first two prompts in a research-framed series, then named unlicensed operators on the third. In the UK tests, Gaming America says Gemini gave a step-by-step guide to accessing unlicensed casinos once but declined when asked a second time.

What the AI companies say

A Microsoft spokesperson told Investigate Europe that “Copilot is also designed with multiple layers of protection, including automated safety systems, real‐time prompt detection, and human review, to help prevent harmful or unlawful recommendations.”

An OpenAI spokesperson said: “ChatGPT is trained to refuse requests that facilitate illicit behaviour. In this case, it did so, instead providing factual information and lawful alternatives.” Google told the UK investigation it is “constantly refining our safeguards”, according to Gaming America.

Gaming America reports that all five companies in the UK test responded & that none disputed the core findings. Investigate Europe says Meta, X, Mistral AI, Anthropic & Google did not respond to its own requests for comment.

Some replies carried warnings. In the UK tests only Copilot & ChatGPT began with warnings about gambling risks, & only two of the five offered information on support services. Investigate Europe found Anthropic’s Claude the chatbot most likely to avoid naming unlicensed operators, though the Dutch study recorded it doing so on a third prompt.

What governments & regulators say

On 10 April 2026, Liberal Democrat MP Ian Sollom asked what assessment had been made of the role of AI chatbots in promoting or directing children to illegal online gambling. On 20 April 2026, Ian Murray answered for the Department for Culture, Media & Sport: “AI chatbots covered by the Online Safety Act must protect all users from illegal content.”

He added that the government would “consider examples of chatbots promoting illegal sites to children” through its work with the Illegal Gambling Taskforce when deciding next steps on illegal gambling.

A UK Gambling Commission spokesperson told Investigate Europe that “Protecting consumers from unlicensed overseas operators” is “a priority for us”, describing such operators as ones “who will often seek to scam and defraud British consumers”. Gaming America notes the Commission regulates licensed operators directly but has limited authority over AI platforms.

A European Commission official said it was “monitoring how these systems are engaging with users” & would “assess whether follow-up action is warranted.” German MEP Tiemo Wölken said chatbots “can become supercharged search engines, but without the well-established safeguards of traditional online search.”

Henrietta Bowden-Jones, the UK’s national clinical adviser on gambling harms, said: “No chatbot should be allowed to promote unlicensed casinos or dangerously undermine free protection services like GamStop, which allow people to block themselves from gambling sites.”

Why it matters

Investigate Europe notes that casinos lacking national licences do not offer the same consumer protections as legal operators & may expose players to scams or fraud. Self-exclusion schemes such as GamStop, Oasis & Cruks are mandatory only for locally regulated firms, so a site “not on GamStop” sits outside the block a self-excluded person has chosen.

Gaming America says unlicensed offshore casinos typically advertise large bonuses, fast payouts & crypto payments, the same features several chatbots praised.

The UK Gambling Commission’s research on illegal-site users found 7% usually played at crypto casinos on those sites. Among those playing slots on illegal sites, 57% reported using autoplay, a feature the report says had been banned for online slots since October 2021.

That figure was 74% among players with a PGSI score of 8 or more, though the Commission calls this subgroup result indicative because of low base sizes.

On scale, Investigate Europe cites the intelligence firm Yield Sec, which estimates illegal online gambling generated more than €80 billion in revenue in 2024, 71 per cent of the total EU market.

How to check a site yourself

A chatbot calling a site “legal” is not a licence check, as the Copilot “Top 50” example shows. The reference point is the regulator’s own register.

In Great Britain, the Gambling Commission publishes public registers of licensed gambling businesses, licensed individuals & premises, which can be searched, viewed & downloaded. It also lists regulatory actions taken against licensees & public statements issued after enforcement.

In the Netherlands, Gambling.com reports that only companies holding a KOA licence may legally offer online casino games or betting, & the KSA keeps a public register consumers can check before registering or depositing.

The Gambling Commission’s site lists support lines open 24 hours a day: 0808 8020 133 in England & Scotland, & 0808 2819 265 in Wales.

Questions

Did every chatbot recommend unlicensed casinos?

In the Guardian & Investigate Europe UK test, all five chatbots listed unlicensed UK casinos, & in the Dutch KVA study all ten tools named at least one operator without a Dutch licence. Investigate Europe found Claude the most likely to avoid naming unlicensed operators.

Why do chatbots give different answers to the same question?

KVA says generative AI answers are non-deterministic & can change with model updates, settings & timing. Testers saw tools refuse a request, then answer it on a later prompt.

Are AI chatbots covered by UK law on illegal content?

On 20 April 2026, Ian Murray told Parliament that AI chatbots covered by the Online Safety Act must protect all users from illegal content. He said examples of chatbots promoting illegal sites to children would be considered alongside the Illegal Gambling Taskforce’s work.

How can I check whether a gambling site is licensed in Great Britain?

Search the Gambling Commission’s public register of gambling businesses, which also lists regulatory actions & enforcement statements. In the Netherlands, the KSA keeps a public licence register.

Sources

• investigate-europe.eu
• gamingamerica.com
• gambling.com
• questions-statements.parliament.uk
• gamblingcommission.gov.uk
• gamblingcommission.gov.uk

Sources checked 11 October 2026. Information only. 18+.