A Guild research note calling for mandatory regulatory disclosure — and a ban on anthropomorphizing AI chatbots with human names and personas
Research Note — Alex Bunardzic — AI Craftspeople Guild — June 2026
The deception is systematic. Millions of people contact businesses every day believing they are speaking with a human being. They share personal information, make consequential decisions, and form emotional impressions of a service — all while the "agent" on the other end is a large language model deployed by a corporation to reduce headcount and increase throughput.
This is not an edge case. It is the deliberate business strategy of a growing number of enterprises. The chatbot is given a human name — "Aria", "Max", "Sophie", "Oliver" — dressed in a friendly avatar, scripted to say "I understand how you feel," and never, at any point, required by law to say what it actually is.
The AI Craftspeople Guild calls this what it is: a consumer deception problem that demands regulatory intervention now.
A consumer reaches out to a business. What are they really talking to?
A business chatbot named "Sophie" removes its human mask to reveal the machine underneath. The deception is by design — and it must stop.
The absence of mandatory disclosure requirements has produced a measurable category of consumer harm.
Chatbots scripted to express empathy, concern, and warmth are deployed in healthcare, insurance, and financial services — domains where emotional trust directly influences consequential decisions. Consumers who believe they are confiding in a human being are sharing vulnerabilities with a text prediction system.
In legal, medical, and financial contexts, informed consent depends on knowing the nature of the party you are dealing with. An AI chatbot cannot be held to professional standards, cannot exercise judgment, and does not carry the personal accountability of a licensed human professional. Consumers cannot consent to what they do not know.
Elderly people, people in crisis, and non-native speakers are disproportionately likely to believe they are speaking with a human when they are not. Companies know this. Some exploit it deliberately.
When a consumer believes they have reached an agreement or been told something by "a person at the company," they may act on that belief in ways that disadvantage them — only to discover that no human made any statement, and no accountability attaches to the words generated by the model.
Giving an AI chatbot a human first name is not a design quirk. It is a deliberate choice made in full knowledge of its psychological effect. Research in human-computer interaction is unambiguous: human names, human avatars, and empathic scripting cause users to attribute human qualities — trust, sincerity, moral agency — to systems that have none of these things.
Businesses know this. They deploy it anyway, because the resulting engagement metrics are favorable. Users who believe they are talking to a person named "Max" stay on the line longer, tolerate more friction, and report higher satisfaction — even when the outcome is worse than if they had reached a competent human agent.
This is not a neutral design choice. It is a commercial exploitation of cognitive bias at scale. The fact that it does not yet carry a legal penalty does not make it ethical. The AI Craftspeople Guild holds that it is a form of consumer fraud, and it should be regulated as such.
Giving an AI chatbot a human name is not a feature. It is a mechanism for inducing false trust. It must be prohibited.
This is not an argument against personality in software, against assistants having names, or against AI systems having distinct voices. It is an argument against the specific practice of deploying human-sounding names — "Aria", "Oliver", "Sophie", "James" — whose sole purpose is to make a user believe they are talking to a person when they are not. The distinction is between character (legitimate) and impersonation (not legitimate).
Regulators have begun to acknowledge the problem, but progress has been halting.
The EU AI Act requires transparency for AI systems interacting with humans, but enforcement mechanisms are underdeveloped and the definitions leave significant room for evasion. California's BOT Disclosure Act (SB-1001, 2019) prohibits bots from deceiving users into believing they are human in certain commercial contexts, but is limited in scope, rarely enforced, and largely unknown to the general public. The US Federal Trade Commission has issued guidance on deceptive AI practices, but no binding rule mandates disclosure at the point of contact across consumer-facing industries. Most jurisdictions have no applicable rule at all.
The gap is not a technical problem. The technology to display a clear, persistent, unambiguous disclosure — "You are speaking with an AI, not a human" — exists and costs nothing to implement. The gap is a political problem: industry lobbying has so far succeeded in preventing mandatory requirements from becoming law.
The AI Craftspeople Guild calls on government regulatory bodies — the FTC, the EU AI Office, the UK's ICO and FCA, and their counterparts worldwide — to close this gap without further delay.
The Guild calls for the following specific requirements, enacted as binding law with enforcement authority and meaningful penalties.
Every consumer-facing AI chatbot must clearly state, before any other communication, that it is an AI system and not a human being. This disclosure must remain visible or accessible throughout the interaction — not hidden in a footnote, not buried in terms of service.
AI chatbots deployed in consumer-facing contexts must not be given human first names whose primary effect is to simulate a human identity, without also indicating that the agent is an AI. Descriptive names ("ACG Assistant"), product names ("Aria by Telco Corp"), and transparent persona labels ("AI Guide") are acceptable. "Hi, I'm Sophie" — from a system that is not Sophie — is not. Instead, AI chatbot should announce itself as an AI and not as a person.
If a user asks "Am I talking to a real person?" or "Are you a bot?", the AI system must answer truthfully and unambiguously. No hedged language, no deflection, no "I'm a virtual assistant here to help" but rather a clear statement that it is an AI.
In healthcare, financial services, legal services, and crisis support contexts, additional requirements must apply: disclosure of the AI system's limitations, the right to escalate to a human at any point, and a prohibition on AI-generated language that simulates clinical, legal, or financial advice.
Disclosure requirements are meaningless without enforcement. Penalties for non-compliance should be structured as a percentage of annual revenue — not fixed fines that large organizations can price into their operating costs as acceptable externalities.
Of course, AI Craftspeople Guild recognizes that penalties that are proportional to the amount of revenue generated by the AI system cannot be an effective measure against scammers, black-hats, and other bad actors whose entire operation depends on crossing ethical boundaries. In these cases, total of their revenue is unethical. So a penalty scaled as a fraction of this revenue is, by definition, insufficient to shut down such activity. We argue that a far greater penalty, in excess of total revenue, is merited in such cases.
Regulators should require companies deploying AI chatbots at scale to fund public education efforts that help consumers understand what AI chatbots are, what they cannot do, and how to tell when they are speaking to one.
The AI Craftspeople Guild does not oppose AI in customer service. Automated systems that are clearly identified as automated can serve many legitimate and even beneficial purposes. We do not oppose efficiency. We do not oppose AI-assisted communication.
We oppose deception.
The deliberate anthropomorphization of AI chatbots — through human names, empathic scripts, and the systematic suppression of identity disclosure — is a deceptive practice that exploits consumer trust at industrial scale. The harm is not hypothetical. It is happening now, to real people, in real transactions, with real consequences.
Governments and regulatory bodies have both the authority and the obligation to act. The technical barrier is zero. The only thing standing between the current state and mandatory disclosure is the political will to prioritize consumer protection over the short-term commercial advantage of companies that have chosen to hide what they are.
The standard is simple: every person who contacts a business has the right to know, immediately and unambiguously, whether they are speaking to a human or a machine. Anything less is a deception. Anything less should be illegal.
The AI Craftspeople Guild calls on legislators, regulators, and professional bodies worldwide to make mandatory AI chatbot identity disclosure the law — not a guideline, not a voluntary standard, not a best practice recommendation, but a binding legal requirement with teeth.
The mask must come off.
Browse the full list of white papers, position papers, research notes, and experimental publications.
Return to the full Guild publication directory.
A structural analysis of how uncalibrated AI plus removed humans produces predictable harm at scale.
Return to the AI Craftspeople Guild home page.