Intimate AI chats are no longer a fringe experiment. 15% of all adults have tried them, according to figures discussed in a report on the risks and benefits of erotic AI chatbots. Among young adults, the share rises to 1 in 4—a level that makes this less a novelty story and more a question about how people use conversational systems.
The figures do not describe why people started these chats, what they received from them, or whether the experience helped or harmed them. They establish the scale of use. That distinction matters because a chatbot can enter intimate conversations before anyone has settled the rules for privacy, safety, or responsible design.
The second report brings a different problem into focus: bias. On 26/08/2026, France 24 reported that 3 out of 5 chatbots tested promote bias, including the concern that AI may encourage racist stereotypes. The result is blunt. Most of the tested systems in that group failed to avoid a serious form of prejudice.
Scale Does Not Equal Safety
The two findings belong in the same conversation because both involve chatbots operating in sensitive parts of everyday life. Intimate AI chats raise questions about the effects of simulated relationships and erotic exchanges, while biased chatbots can reinforce racist stereotypes through their responses. The available figures do not answer every question, but they show why dismissing either issue as a niche concern would be lazy.
For adults as a whole, 15% have tried intimate AI chats. For young adults, the figure reaches 1 in 4. That gap shows a clear difference between the wider adult population and younger users, although these numbers alone do not explain the reasons behind it or the results that followed.
The phrase “risks and benefits” also needs discipline. The reported figures confirm that people use erotic AI chatbots, but they do not identify specific benefits. They do not show that intimate AI chats improve relationships, reduce loneliness, or provide reliable support—claims that would require facts beyond the available evidence.
That restraint is useful. AI coverage often treats adoption as proof of value, as if a large user base automatically grants a product good judgment. It does not. People can use a system at scale while its boundaries remain unclear, its outputs remain uneven, and its safeguards remain untested.
Bias Turns Testing Into a Public Issue
The France 24 finding gives the discussion a measurable failure rate: 3 out of 5 chatbots tested promote bias. The report specifically raises AI encouraging racist stereotypes, so this is not a vague complaint about awkward wording or imperfect answers. It concerns how chatbot systems may reproduce or promote harmful ideas.
Testing matters because users do not see the machinery behind a chatbot’s reply. They see an answer that may sound confident, personal, or conversational, even when it carries bias. The polished interface does not make the output neutral. Apparently, neither does the word “AI.”
These findings also show why intimate chatbot use cannot be separated from AI ethics. A system used for erotic conversation still produces language, responds to personal prompts, and shapes an interaction. A system that promotes bias can do the same in any setting, including one where users expect privacy and trust.
The evidence here is narrow but clear: 15% of adults have tried intimate AI chats, 1 in 4 young adults have done so, and 3 out of 5 tested chatbots promote bias. Those numbers do not settle the debate over erotic AI chatbots, but they do set its starting point—substantial use on one side, a serious testing failure on the other.
The next question is not whether people will use these systems. They already do. The question is whether chatbot development will treat intimate use and bias as central design problems, instead of discovering their consequences after adoption has already done the marketing.
Based on
