AI Chat Agent and Chatbot Statistics 2026: Market Size, Adoption, What Customers Want, Cost per Resolution and the Law, Every Number Sourced

In short: The global chatbot market is estimated at $10 billion to $12 billion for 2026, depending on the research firm. [1][2] In Germany, Bitkom reports that for the first time a majority of companies with 20 or more employees use AI, most often in customer contact. [11] Customers use AI in their private lives, but they do not use company chatbots any more than they did in 2022, and 87 per cent say access to a human is essential. [8][20] If you introduce an AI chat agent, keep three things in view: how many requests it really resolves, how quickly it hands over to a person and whether it identifies itself as AI.

We rewrote this article from scratch in October 2026. The previous version promised more than 80 figures. Around 40 of them could not be backed up: conversion uplifts, payback periods, industry rates and studies we could not find at the publishers named. They have been removed. Statements about our own product were also wrong. The EchoCall chat agent does not run on WhatsApp, Telegram, SMS or Facebook Messenger. It runs on your website. It speaks more than 30 languages, not the much higher number that appeared here before. The smallest chat plan costs €40 a month, not €49. We have also deleted the claims about worldwide hosting, about EchoCall leading the market and about EchoCall case studies. Every figure on this page now carries a number in square brackets that leads to its source at the end. Where a figure is only a vendor's own claim or a survey run by a software vendor, we say so.

You are welcome to quote any figure. Please cite the original source, not just us.

The key figures at a glance

1. Market size: how big is the chatbot market in 2026?

Anyone searching for the size of the chatbot market finds figures that differ several times over. The reason is almost always the definition. Some reports count only chatbot platforms and the services around them. Others count "conversational AI", which also covers voicebots, virtual assistants and GenAI agents for text and speech. If you quote a figure, say which market it refers to.

For 2026 the two chatbot reports are close: $10.42 billion at Fortune Business Insights, $11.8 billion at Grand View Research. [1][2] For 2033 and 2034 they drift apart, because five percentage points of annual growth add up over eight years. According to Grand View Research, customer service was the largest application in 2025. [2] On the same page, MarketsandMarkets gives 2030 in one place and 2031 in another as the target year for the $49.80 billion. We give both because we do not know which is correct. [3]

Gartner does not publish a market size for chatbots. The Gartner figure most often quoted in this context dates from August 2022: conversational AI would cut contact centre agent labour costs by $80 billion in 2026. Gartner forecast spending on conversational AI in contact centres at $1.99 billion for 2022. At the time, Gartner estimated that 1.6 per cent of agent interactions were automated and expected one in ten by 2026. [4] Whether that happened, Gartner has not measured in the releases we found.

What the figures measure: vendor revenue, estimated by research firms that sell their full reports. What they do not measure: how many requests chatbots actually resolve, or whether customers are more satisfied. For your decision, the market size tells you mainly one thing. There are many providers, and prices are moving.

2. Consumer use: AI yes, company chatbots not so much

Eurostat measured the use of generative AI for the first time for 2025. In the EU, 32.66 per cent of people aged 16 to 74 had used it in the three months before the survey, in Germany 32.25 per cent. In Germany, 27.2 per cent used it privately and 15.79 per cent for work. [5]

The ARD/ZDF media study 2026 arrives at almost twice that figure. 59 per cent of people aged 14 and over use AI tools at least occasionally, 92 per cent of those aged 14 to 29 and 15 per cent of those aged 70 and over. Weekly use has almost doubled compared with the previous year. At the same time, 54 per cent express scepticism. [6]

Both figures are right. They simply measure different things. Eurostat asks about the last three months and describes 2025. ARD/ZDF asks about occasional use, includes 14 and 15 year olds as well as people over 74, and surveyed 2,462 people between late January and mid April 2026. More than a year lies between the two, and use changed quickly in that time.

For chat agents in customer service, another figure matters more. In February and March 2026, Gartner surveyed 3,566 B2B and B2C customers. The result: for service problems, customers are about three times more likely to turn to third-party AI tools than to the company's own chatbot. Use of third-party GenAI almost doubled in a year. Use of company chatbots has been statistically unchanged since 2022. [8] Already in 2025, according to Gartner, more than half of all customer service journeys (51 per cent) began on third-party platforms such as Google, YouTube or ChatGPT, and 74 per cent for Generation Z. Only 22 per cent of customers resolved their issue entirely within the company's own channels. [7]

Gartner gives the reason as well. Many company chatbots are still built to answer questions. Customers increasingly expect AI to do things for them: book an appointment, submit documents, update an account. Among customers who use GenAI, 58 per cent have had it complete a task on their behalf, 74 per cent in B2B. [8]

What this means for you: a chat agent that only repeats your FAQ competes with ChatGPT and loses. An agent that books an appointment, looks up an order or creates a ticket can do something the external tool cannot, because that tool has no access to your systems. Our article AI Agents vs. Chatbots: What's the Difference? (2026) explains the difference between the two.

3. Business adoption: how many companies use AI in customer service

For Germany there are two official sources and one industry association survey. They reach different results, and there are reasons for that.

Source: Eurostat, dataset isoc_eb_ai, updated 15 June 2026 [9]

Eurostat. In 2025, one in five EU enterprises with ten or more employees used AI, up 6.47 percentage points in a year. Among large enterprises it was 55.03 per cent, among medium-sized ones 30.36 per cent and among small ones 17 per cent. Denmark (42.03 per cent), Finland and Sweden lead. Romania (5.21 per cent), Poland and Bulgaria come last. At just under 26 per cent, Germany is above the EU average. [9]

Destatis. The Federal Statistical Office arrives at the same figure for Germany, 26 per cent. For companies with 10 to 49 employees it is 23 per cent, with 50 to 249 employees 36 per cent, with 250 or more 57 per cent. Among companies that use AI, 52 per cent use text mining, 42 per cent speech recognition and 35 per cent natural language generation, which Destatis takes to include program code. From 2025 the EU definition of an enterprise applies, so comparisons with 2024 are only possible to a limited extent. [10]

Bitkom. In September 2026 the German digital industry association reported that for the first time 57 per cent of companies use AI, after 36 per cent a year earlier and 20 per cent in 2024. AI is most often used in customer contact, for example in handling enquiries (72 per cent), ahead of marketing and communication (54 per cent). Bitkom surveyed 603 companies with 20 or more employees by telephone in July and August 2026. The release does not state the base for the 72 per cent explicitly. From the context, it refers to companies that use AI. [11]

Why 26 and 57 per cent do not have to match: Destatis and Eurostat collect official data from all enterprises with ten or more employees and describe 2025. Bitkom surveys 603 companies with 20 or more employees by telephone in summer 2026. The smallest businesses, which use AI least, are not included by Bitkom, and a year lies between the surveys. If you need one figure for "Germany", take the official one and state the year. If you want to know where the trend is heading, look at Bitkom.

4. Customer service metrics: how many requests AI chat agents really resolve

The resolution rate is the metric almost everyone asks about. It is also the hardest to compare, because every vendor defines "resolved" differently.

Why 70 to 76 per cent and one interaction in ten can both be true: the vendor rates refer to conversations that reach the bot, in companies that chose the product and look after it. Gartner counts all interactions with service agents, including phone and email, and including companies without a bot. None of the vendors discloses how its rate is calculated. Whether a customer who simply gives up after the bot's answer counts as resolved depends on the definition. At Intercom, for example, an outcome also counts if the customer does not follow up after Fin's answer. [12]

The Klarna example. In February 2024 Klarna reported that its AI assistant had held 2.3 million conversations in its first month, two thirds of all service chats. It was doing the work of 700 full-time agents, resolving issues in under 2 minutes instead of 11, with 25 per cent fewer repeat enquiries. Klarna estimated the profit improvement for 2024 at $40 million. [14] All of these are the company's own figures. In May 2025 Bloomberg reported that chief executive Sebastian Siemiatkowski believed AI-driven cost cutting in customer service "has gone too far". Klarna wanted to recruit people again so that customers "will always have the option of speaking to a real person". [15] We did not find a Klarna release on this, so the source here is the press.

What managers report. Gartner expects that by 2027 half of the organisations that planned to cut their customer service workforce significantly because of AI will abandon those plans. In a survey of 163 service leaders in March 2025, 95 per cent intended to keep human agents. In October 2025 only 20 per cent of 321 respondents had reduced staff because of AI, and 55 per cent kept headcount stable. [17] In a second Gartner survey of 1,303 leaders, the median share of the 2025 budget that went into AI was 12 per cent. Only 24 per cent could show a positive financial return. [8]

What to measure yourself rather than trusting other people's rates:

• the share of conversations that end without a handover and after which the customer does not come back with the same question within a few days,
• the share of handovers to your team and the time until a person takes over,
• the questions the agent could not answer. They show what is missing from the knowledge base.

5. Expectations and trust: what customers expect from chat

The most precise figure for Germany comes from Bitkom. Online shoppers were asked whom they would like to contact if something goes wrong with a purchase, and how satisfied they were with the contact channels they had already used.

Source: Bitkom, 1,006 respondents aged 16 and over, 978 of them online shoppers, January 2025, representative [18]

The chatbot sits mid-table for preference and almost at the bottom for satisfaction. Half of its users were satisfied, the other half were not. That is not a verdict on AI as such. The question covered every chatbot online shoppers had come across by early 2025, including many rule-based systems.

Gartner has surveyed customers about AI in service twice, a little over two years apart. In December 2023, 64 per cent of 5,728 customers said they would prefer companies not to use AI in customer service. 53 per cent would consider switching to a competitor. The biggest concern was that it would become harder to reach a person, ahead of job losses and wrong answers. [19] In spring 2026, 50 per cent said their interactions with a company had become easier with GenAI. But 87 per cent said it was essential to have the option of reaching a human agent. Customers who did not want to deal with AI most often named one thing that would change their mind: the option to switch to a person when needed. [20]

Between 64 per cent rejection and 50 per cent "easier" lie a little over two years and a different question. The common ground is clear: customers do not reject the AI, they reject the locked door behind it.

Two vendor studies point in the same direction. They come from software companies that sell AI for customer service. That does not make them wrong, but you should know it.

• Zendesk, CX Trends 2026: 74 per cent of consumers say that because of AI they now expect customer service to be available around the clock. 95 per cent expect an explanation for decisions made by an AI, and only 37 per cent of companies provide one. According to Zendesk, 6,182 consumers and 5,115 CX professionals in 22 countries were surveyed. Fieldwork dates and the publication date are not on the page. [21]
• Salesforce, State of the AI Connected Customer: 72 per cent of customers say it is important to know whether they are communicating with an AI agent. Sample according to Salesforce: 16,585 consumers and business buyers, with no fieldwork dates. [22]

6. Germany and the EU: what has applied since 2 August 2026

Article 50 of the AI Act has applied since 2 August 2026. Providers must design AI systems intended to interact directly with people so that those people are informed that they are interacting with an AI system. This must happen in a clear and distinguishable manner at the latest at the time of the first interaction. The only exception is where this is obvious anyway to a reasonably well-informed, observant and circumspect person. Breaches of the transparency obligations under Article 50 can cost up to €15 million or up to 3 per cent of worldwide annual turnover, whichever is higher. [23] The European Commission published guidelines on these obligations on 20 July 2026. [24]

In summer 2026 the so-called Digital Omnibus on the AI Act was adopted, Regulation (EU) 2026/1744, in force since 27 July 2026. It gives systems that were already on the market before 2 August 2026 until 2 December 2026 to comply with the machine-readable marking of AI-generated content under Article 50(2). It neither changed nor postponed the "you are talking to an AI" notice under Article 50(1). [25] If you read that the AI disclosure duty for chatbots has been postponed, the source meant a different provision.

One question remains open. The wording of Article 50(1) names the providers. Whether a company that buys a ready-made chat agent and puts it on its website is itself obliged is a matter of interpretation. In practice it hardly matters, because the notice costs one sentence. In the EchoCall chat widget, the system shows the AI notice automatically, in German, English and French. All the details are in EU AI Act 2026: What the AI Disclosure Duty Means for Voice and Chat Systems.

7. WhatsApp chatbots in Germany: use, prices, rules

We did not find an official figure for WhatsApp alone. Eurostat measures instant messaging as a whole, meaning services such as WhatsApp, Skype or Viber. In 2025, 82.79 per cent of people aged 16 to 74 in Germany used them, and 82.31 per cent in the EU. [26]

Costs have just changed. Since 1 July 2025 Meta has charged per delivered template message, meaning marketing messages, utility messages such as shipping updates, and authentication codes. Free-form replies in the open 24-hour window after a customer message were free. Since 1 October 2026 Meta also charges for these service messages per message, at the same rates as utility and authentication messages in each market. Each business phone number has a free tier of 1,000 delivered service messages a month, and unused messages expire at the end of the month. [27] Meta publishes the price per message for Germany only in downloadable rate cards, which we have not evaluated. Calculations that assume replies to customer questions cost nothing have been out of date since October 2026.

Since 15 January 2026, and for new customers since 15 October 2025, Meta's terms have prohibited providers of large language models and general-purpose AI assistants from using the WhatsApp Business Platform where the AI is the primary functionality rather than an incidental or ancillary one. Going by the wording, a service bot for your own business is not covered. Meta allows exceptions for certain countries. The current version of 23 September 2026 refers to a separate Meta page for these. [28] How to set up WhatsApp with AI properly is covered in WhatsApp Business with AI in 2026: What Meta Allows, What It Costs and How to Stay GDPR Compliant.

For EchoCall: WhatsApp is currently not part of what you set up yourself in the EchoHub. Today the EchoCall chat agent runs on your website, with handover to your team.

8. Costs: what an AI chat agent costs per conversation and per resolution

The big help desk vendors no longer charge for AI agents per seat but per outcome. Each of them defines an outcome in its own way, so the prices are not directly comparable.

What counts as resolved is in the small print. At HubSpot a conversation is resolved if it is not handed to a person within 72 hours. [13] At Zendesk a resolution counts if the issue is resolved without an agent stepping in, and for email only after 72 hours without a further customer reply. Depending on the plan, 5 to 15 automated resolutions per agent and month are included. [29] At Intercom an outcome counts if the customer confirms the solution, does not follow up after Fin's answer or Fin completes a workflow, which according to Intercom includes a handover. Each conversation is charged at most once. On top come Intercom seats at $29 to $132 per seat and month on annual billing. [12] Salesforce keeps the $2 conversation price, alongside actions at $0.10 and voice actions at $0.15. [30] The Freshworks page does not make clear whether the 500 included sessions apply every month or only once. [31]

Worked example. Assumptions: 1,000 chat conversations a month, of which 600 count as resolved under each vendor's definition. The 60 per cent is an assumption for the calculation, not a measured value. We leave out seats, base fees, included allowances and tax.

The calculation does not show which product is cheaper. The help desk vendors sell their AI agent as part of a suite with ticketing, seats and reporting. It shows something else: the billing model shapes the bill more than the unit price does. If you pay per resolution, you pay less when the rate is poor, but you pay for everything the vendor counts as resolved. If you pay per conversation, you know in advance what the month will cost, but you carry the risk of a weak agent yourself.

EchoCall in brief. The chat agent costs €40 a month for 100 chat conversations, €120 for 300, €400 for 1,000 and €1,000 for 5,000 or more, all excluding VAT. [32] It runs on any website via a script snippet or via the WordPress plugin, speaks more than 30 languages and answers from your knowledge base, for example from PDFs, individual pages or your whole website. Your team can take over conversations in the Live Inbox. Since October 2026 visitors can also talk to the agent by voice in the widget, with the minutes coming from the voice allowance. Before the microphone starts, a notice card appears. For testing, there are 30 voice minutes and 30 chat conversations for 14 days, and the credit card is verified with €1. Without signing up, you can enter your website address in the live demo and after about 20 seconds chat with an agent built from your public pages.

9. When chatbots get it wrong: four documented cases

Statistics show averages. The following cases show what happens in a single case when a chatbot works without limits.

• Air Canada, Canada, February 2024. The airline's chatbot had given a customer wrong information. Air Canada argued that the chatbot was "a separate legal entity". The tribunal in British Columbia called this a "remarkable submission". It should be obvious to Air Canada that it is responsible for all the information on its website. Air Canada had to pay CA$650.88 in damages, CA$812.02 including interest and fees. [33]
• DPD, United Kingdom, January 2024. After a system update, a customer got the chatbot to swear and to write a poem about DPD. According to the BBC, his post was viewed 800,000 times in 24 hours. DPD immediately disabled the AI element. [34]
• Cursor, USA, April 2025. The software company's support bot invented a rule that a subscription only works on one device. The co-founder apologised publicly. Since then, AI responses in email support have been clearly labelled as such. [35]
• New York City, March 2024. According to The Markup, the city's AI chatbot for business owners claimed that bosses could take their workers' tips and that landlords could discriminate based on source of income. Both are wrong. [36]

The lesson from all four cases: responsibility stays with the company, whoever answers. A chat agent needs a maintained knowledge base, clear limits on topics it should not talk about and a handover to people before things get embarrassing.

10. What the figures do not tell you

• No independent resolution rates. All rates in this article come from vendors or from the company itself. We did not find an independent, publicly verifiable measurement.
• No reliable conversion figure. The previous version gave several values for how much a chat increases a website's sales. We could not find a primary source for them. If you want to know, measure it yourself, for example with a test with and without the widget.
• No official figure on company chatbots. Eurostat and Destatis measure AI use in general. Nobody officially records how many companies in Germany have a chatbot on their website.
• Market sizes are estimates. The methodology sits in reports that cost money. The public summaries are enough to know the order of magnitude, not for a purchasing decision.
• Vendor surveys. Zendesk and Salesforce state neither fieldwork dates nor publication dates on their pages. We quote them because they have large samples, and we label them.

Frequently asked questions about AI chat agent statistics

How big is the chatbot market in 2026?

Between $10.42 billion and $11.8 billion, depending on the research firm. Fortune Business Insights expects $60.21 billion by 2034, Grand View Research $41.2 billion by 2033. The broader conversational AI market, which also includes voice systems, stood at $17.05 billion in 2025 according to MarketsandMarkets.

How many companies in Germany use AI in customer service?

In 2025, according to Destatis, 26 per cent of all enterprises with ten or more employees used AI. Among companies with 20 or more employees, Bitkom put the figure at 57 per cent in summer 2026, and 72 per cent used AI in customer contact, for example in handling enquiries. There is no official figure for chatbots in customer service alone.

How many requests can an AI chat agent resolve?

Vendors quote 70 to 76 per cent on average across their customers, which has not been independently verified. Gartner estimated that only 1.6 per cent of all agent interactions were automated in 2022 and expects 80 per cent of common issues to be resolved without humans only by 2029. The rate in your company depends on your requests and your knowledge base. Measure it yourself in the first few weeks.

Do customers actually want to chat with a chatbot?

For simple questions yes, as the only option no. 36 per cent of German online shoppers want a chatbot when something goes wrong, 62 per cent a person they can reach quickly. In a Gartner survey, 87 per cent of customers said access to a human is essential when a company uses AI in service.

What does an AI chat agent cost?

The big help desk vendors charge per resolution: $0.50 at HubSpot, $0.99 at Intercom, €1.50 to €2 at Zendesk, each on top of the licences. Salesforce lists $2 per conversation. At EchoCall, 100 chat conversations cost €40 a month and 5,000 conversations €1,000.

Does a chatbot have to say that it is an AI?

Yes. Since 2 August 2026, Article 50 of the AI Act has required that people learn they are dealing with an AI at the latest at the first interaction, unless it is obvious. The Digital Omnibus did not postpone this. Breaches can cost up to €15 million or 3 per cent of worldwide annual turnover.

Does an AI chat agent also work on WhatsApp?

Technically yes, via the WhatsApp Business Platform. Meta's terms only prohibit AI assistants where the AI itself is the product. Since 1 October 2026, however, Meta also charges for service messages once the 1,000 free messages per number and month have been used. WhatsApp is currently not part of what you set up yourself in the EchoHub. Today the EchoCall chat agent runs on your website, with handover to your team.

Sources

1. Fortune Business Insights, Chatbot Market Size, report FBI104673, last updated 21 September 2026 ($8.37bn in 2025, $10.42bn in 2026, $60.21bn in 2034): fortunebusinessinsights.com
2. Grand View Research, Chatbot Market Size Report, June 2026 ($9.6bn in 2025, $11.8bn in 2026, $41.2bn in 2033): grandviewresearch.com
3. MarketsandMarkets, Conversational AI Market, May 2025 ($17.05bn in 2025, $49.80bn for 2030 or 2031): marketsandmarkets.com
4. Gartner, press release of 31 August 2022, "Gartner Predicts Conversational AI Will Reduce Contact Center Agent Labor Costs by $80 Billion in 2026": gartner.com
5. Eurostat, Digital economy and society statistics, households and individuals, and dataset isoc_ai_iaiu (use of generative AI in 2025): Statistics Explained, isoc_ai_iaiu
6. ZDF, press release of 22 September 2026, "ARD/ZDF-Medienstudie 2026: Nutzung von KI-Tools nimmt Fahrt auf", and Media Perspektiven 25/2026 (both in German): presseportal.zdf.de, media-perspektiven.de
7. Gartner, press release of 23 July 2025, "Gartner Survey Finds More Than Half of Customer Service Journeys Now Start on Third-Party Platforms": gartner.com
8. Gartner, press release of 8 July 2026, "Gartner Survey Finds Customers Are 3x More Likely to Use Third-Party GenAI Than Company-Provided Chatbots for Customer Service": gartner.com
9. Eurostat, Use of artificial intelligence in enterprises, and dataset isoc_eb_ai: Statistics Explained, isoc_eb_ai
10. Federal Statistical Office (Destatis), table "Unternehmen, die künstliche Intelligenz nutzen", as of 24 November 2025 (in German): destatis.de
11. Bitkom, press release "Erstmals nutzt die Mehrheit der Unternehmen KI" of 14 September 2026, 603 companies with 20 or more employees (in German): bitkom.org
12. Intercom, Fin: home page (76% average resolution rate), pricing ($0.99 per outcome) and seat pricing: fin.ai, fin.ai/pricing, intercom.com/pricing
13. HubSpot, Customer Agent (70% average resolution rate, $0.50 per resolution): hubspot.com
14. Klarna, press release of 27 February 2024, "Klarna AI assistant handles two-thirds of customer service chats in its first month": klarna.com
15. Bloomberg, 8 May 2025, "Klarna Slows AI-Driven Job Cuts With Call for Real People" (paywall): bloomberg.com
16. Gartner, press release of 5 March 2025, "Gartner Predicts Agentic AI Will Autonomously Resolve 80% of Common Customer Service Issues Without Human Intervention by 2029": gartner.com
17. Gartner, press releases of 10 June 2025 (workforce reduction plans) and 2 December 2025 (only 20% have reduced staff because of AI): gartner.com, 10 June 2025, gartner.com, 2 December 2025
18. Bitkom, press release "Kundenservice beim Online-Shopping: Mensch schlägt Chatbot" of 22 May 2025 (in German): bitkom.org
19. Gartner, press release of 9 July 2024, "Gartner Survey Finds 64% of Customers Would Prefer That Companies Didn't Use AI for Customer Service": gartner.com
20. Gartner, press release of 4 August 2026, "Gartner Survey Finds 87% of Customers Say Companies Using GenAI for Customer Service Must Provide Access to a Human Agent": gartner.com
21. Zendesk, CX Trends 2026 (vendor study): cxtrends.zendesk.com, methodology
22. Salesforce, State of the AI Connected Customer (vendor study): salesforce.com
23. Regulation (EU) 2024/1689 (AI Act), Articles 50, 99 and 113: eur-lex.europa.eu
24. European Commission, 20 July 2026, "Commission publishes guidelines on transparency obligations for providers and deployers of certain AI systems": digital-strategy.ec.europa.eu
25. Regulation (EU) 2026/1744 (Digital Omnibus on the AI Act), Official Journal of 24 July 2026: eur-lex.europa.eu
26. Eurostat, dataset isoc_ci_ac_i, internet activities of individuals in 2025 (instant messaging): ec.europa.eu
27. Meta for Developers, Pricing on the WhatsApp Business Platform (as of 30 September 2026): developers.facebook.com
28. Meta, Meta Terms for WhatsApp Business Platform, section 4.7, version of 23 September 2026, and the earlier version as WhatsApp Business Solution Terms: facebook.com, whatsapp.com
29. Zendesk, pricing (in German) and help article "About automated resolutions for AI agents": zendesk.de/pricing, support.zendesk.com
30. Salesforce, Agentforce: press releases of 12 September 2024 and 15 May 2025, pricing page: salesforce.com, 2024, salesforce.com, 2025, salesforce.com/agentforce/pricing
31. Freshworks, Freshdesk pricing: freshworks.com
32. EchoCall pricing: echocall.de/en/pricing
33. Civil Resolution Tribunal of British Columbia, Moffatt v. Air Canada, 2024 BCCRT 149, decision of 14 February 2024: decisions.civilresolutionbc.ca
34. BBC News, 19 January 2024, "DPD error caused chatbot to swear at customer": bbc.com
35. Ars Technica, 17 April 2025, "Company apologizes after AI support agent invents policy that causes user uproar": arstechnica.com
36. The Markup, 29 March 2024, "NYC's AI Chatbot Tells Businesses to Break the Law": themarkup.org

All links were checked on 6 October 2026. This post is not legal advice.

Read next

• AI Voice Agent Statistics 2026: Market Size, Adoption, Cost per Minute, Highest-Value Niches and the Law, Every Number Sourced
• WhatsApp Business with AI in 2026: What Meta Allows, What It Costs and How to Stay GDPR Compliant
• EU AI Act 2026: What the AI Disclosure Duty Means for Voice and Chat Systems
• ChatGPT in Customer Service 2026: What Works, Where It Gets Risky, and When You Need Your Own AI Agent
• AI chat agent: features and pricing