Last week, Mark Sewards, Leeds South West and Morley’s freshman MP, announced that he had created an AI chatbot version of himself, complete with a facsimile of his voice and an uncanny avatar. While Sewards has become the first MP to take such a step, this is not the first time that Neural Voice, the tech company behind the chatbot, has dabbled in politics; in 2024, they fielded an AI version of their chairman Steve Endacott as an Independent candidate in the Brighton Pavillion.
The West Yorkshire Labour MP said that his chatbot will “help strengthen the connection between an MP’s office and the constituents we serve” by allowing people to ask for help with local issues or policy queries and providing access to that support “24/7, 365 days a year”. However, this will likely have the opposite effect.
Prof Victoria Honeyman, a British politics lecturer at the University of Leeds, gave a nuanced verdict on Sewards’ chatbot. She said that if used to “answer simple messages, then most people would be relatively comfortable as we have [chatbots] in lots of different areas of our lives nowadays”, thus granting Sewards more time to focus on complicated casework. However, she conceded that it “might cause more upset” and “[undermine] people’s confidence in their MP” if mistakes are made when contending with more complicated, potentially emotionally challenging cases.
With Sewards admitting that his new chatbot is a “prototype”, he acknowledges that adjustments may be needed. It would be unfortunate if such adjustments were necessary as a result of of serious mistakes made by the AI that will negatively affect inquiring constituents. For a real-life example of AI failing under such circumstances, last year a bereaved Air Canada passenger (flying to attend his grandmother’s funeral) was misdirected by the chatbot to purchase a full price ticket rather than a bereavement discount ticket; having been told by the chatbot that he would be reimbursed the difference, Air Canada refused it. While this was an embarrassment for a private company, a community’s champion at Westminster making such mistakes would be a dereliction of duty.
Many aspects of our parliamentary democracy make millions of people feel ignored by or alienated from decision makers, or indeed their own MPs. First Past the Post grants diverse communities a single representative who does not need to secure a simple majority of votes to win, thus leading to grossly unrepresentative Parliaments and Governments with undue majorities. It has been long debated whether, in theory, an MP should act only in accordance with the views of their constituents or exercise independent judgement in deciding how best to serve their seat, regardless of popularity. However, with both the traditional whip system and a recent political culture of treating or misrepresenting the Prime Minister as a presidential role, MPs of the two major parties can be expected to serve as lobby fodder (even prioritising their party at the expense of their constituents) or points towards securing the premiership.
Sewards has compared his chatbot to the innovation of contacting your MP via email, but this is arguably a false one; with email, even with standard replies and the right for MPs not to respond, there is the expectation that concerns are being evaluated by a person. Prof Honeyman has said that this will not end in-person interactions such as surgeries; however, there is the possibility that the mechanism for answering simple questions could be used to automatically deflect concerns raised by constituents that the MP or their party would not wish to address. Its provision of information about party policy positions could prove helpful to constituents, but would the governing party consider this helpful for them? This chatbot, in its current form, may ultimately place another barrier between the people and their representatives.
Employing AI like this may be dangerous. While many find virtual assistants such as Alexa or Siri convenient, others are distressed by implicit privacy concerns, primarily the collection of personal data (by monitoring conversation for cues) that may be used for AI training or marketing. Although clear self-identification is a prerequisite to receiving a response from your MP via letter or email, constituents’ private information concerning matters including housing, immigration status and healthcare should not be subject to such jeopardy.
Furthermore, Sewards’ chatbot may contribute to AI’s undermining of what is objective reality in a ‘post-truth’ political environment, beyond mistaking the AI for Sewards himself. It has already been noted for contradicting his voting record on benefits; although this was likely an accident, there is potential for constituents to be misinformed about how their MPs are acting on their behalf. With AI-generated content misleading the public about politicians, wars and natural disasters (to the point of hampering first response efforts), and some public-access AIs being susceptible to public episodes of sabotage, the granting of public access to a programmable facsimile of Sewards’ voice may give committed bad actors opportunities for mischief. Are these good trade-offs for convenient round-the-clock access?
* Samuel James Jackson is the Chair of the Policy Committee of the Yorkshire and the Humber Liberal Democrats and had served as the Liberal Democratic candidate in Halifax during the 2024 general election.



5 Comments
We shouldn’t perhaps be too smug about “MPs of the two major parties”. Christine Jardine wasn’t exactly supported for exercising her independent judgement.
Thanks Samuel, and I agree that this is a terrible move for all the reasons you mention.
I would add that it must also be a privacy and GDPR minefield, with constituent’s voice data sent to the US for processing.
This reads like an article where the author doesnt like AI and is trying to come up with reasons to oppose its use. Can it make mistake – sure . doesa it need checking – of course. But AI would be (if properly managed ) very useful in helping MPs staff deal with the mass of casework MPs get as well as all the email like ‘ please tell me your views on the XX Bill’.
@Simon McGrath
I would not be so starry-eyed about AI as you seem to be. This article from Quanta magazine shows that it takes very little to turn an AI response to the dark side. If we have to check everything AI writes before it is sent out, we might as well write it ourselves. Most responses I get from my MP are pretty formulaic so I doubt an AI response would be very different.
https://www.quantamagazine.org/the-ai-was-fed-sloppy-code-it-turned-into-something-evil-20250813/
“freshman MP” – ugh