When is a bot not a bot? When it’s an Abba concert-performing avatar and marketed as such? Or when it’s clear you are engaging with a human being in real-time?
What if the distinction between the two becomes increasingly opaque? Who has the final word? And who is responsible (and legally accountable) for what happens next? You, or the Chatbot? Constituents in Leeds South West and Morley may be about to find out.
An AI bot version of Mark Sewards MP was unveiled at the start of August. The Leeds-based startup Neural Voice developed ‘the UK’s first virtual MP’. Using recordings of his authentic voice, intonation, mannerisms, style, and image, it created a digital bot mimicking him to interact online with constituents (or perhaps whoever else logs on) 24/7/365.
Stand and deliver: deception by design?
Chatbots like this are usually sold to the public as a convenience gain – and to buyers as a cost-cutting efficiency gain – allowing people to do two or three things at once. Social media algorithms pull us in with these interactions and we can even be lured into thinking that a human is responding when they are not.
Most of us have experienced the ups and downs of chatbots, including the government’s favourite AI health bots. Many of us may be persuaded by the alleged benefits of age-verification tools and assume these tools are accurate, immune from manipulation and can be trusted and relied on, especially if a friendly robot ‘chats’ with us.
Yet, we know that the digital world is hackable, produces what is called ‘slop’, and creates false information and disinformation. AI has the potential to become self-learning and ‘autonomous’, meaning that a string of code powering one bot may be used by that self-taught bot to instruct another bot, or entire sets of bots, in order to create havoc, interfere with infrastructure, or do something dangerous or destructive.
More devastating scenarios may be foreseeable if Generative AI is connected and left to complete tasks for humans. Open AI’s CEO Sam Altman, along with experts and Nobel Prize winners, has repeatedly underlined the risks – even of human extinction – and cautioned that humans must stay in control of AI, and curb bot-to-bot autonomy. However, is that happening?
Losing control to AI
Google’s Gemini AI was given a ‘poisoned calendar invitation’ as a test to take over a smart home. A fully automated hacking system, gpt-oss, is more than a twinkle in the eye. AI hacking systems can search for and patch vulnerabilities in computer systems. But, like all technologies, they can be used for good or ill.
Altman has suggested that an AI that can stand in for authentic humans, using dynamic learning loops to create super artificial intelligence by 2030, is more than a pipe-dream. OpenAI’s new, most advanced model, GPT-5, is supposedly safer, self-learning and geared to what the user wants. It’s not inconceivable that eventually it could be ‘super intelligent’ and put chatbots like Mark Seward’s in the shade.
With the race to develop superintelligent AI accelerating, commercial considerations seem to top protecting humans against potentially malevolent AI or inadvertent harms. As is already clear, many of the greatest advances, which would most benefit physically or mentally disadvantaged or impaired people, are rarely universally accessible to them.
Moreover, dependence on humanoid assistants, brain implants or exoskeletons, for example, can be technically facilitated. Yet, inadequate resourcing for breakdowns and technical failure (battery failures, hacking and so on) risks leaving the vulnerable, dependant users stranded.
The ethical gap
Humans get emotionally attached to their humanoid AI ‘companions’, whether they take the form of a robot animal ‘helper’ with big baby eyes, a sexual companion, or an online pal. Such a response helps to ensure repeated use – up to a point.
At what point can a human user refuse to engage with a bot and still get a service? Will the MP’s bot-assistant ‘resign’ if its self-learning loops suggest lies that would satisfy dissatisfied constituents? Or will MPs using them blame the bot for unpopular decisions come election time?
Civic AI
A lot of research is going into trying to embed ethical values, as broadly understood in Europe, to promote trustworthy AI that prioritises human engagement.
Among them, the AI4Trust project, aims to enhance work done by humans by creating a trust-based environment that integrates the automated monitoring of social and news media with advanced AI-based technologies. It uses AI to combat misinformation and disinformation.
At the level of civil society, a new initiative – Civic AI – has been launched, designed to prioritise AI for the benefit of society, over commercial or opaque government goals.
Media innovator, Christophe Leclerq, founder of Euractiv and Europe Media Lab, explained that Europe must consider buildingTrusted European Platforms (TEPs) as the way to tackle toxic social media networks and reimagine the digital public square. Such TEPs must serve us all and not be controlled by any single, stakeholder community. In short, they must foster trust and collaboration, not surveillance, for the sake of human well-being.
Bye-bye chatbot?
Such a scenario would put the idea of AI being there to serve civil society centre stage. Could that help us all to rethink the purpose of chatbots like Seward’s virtual MP assistant? Will we, or MPs, blame the bots for unpopular political answers or policies? Or will we decide not to bother to turn out to vote come the next election if our MPs delegate answering questions to potentially hackable bots doing unclear things with our personal information?
For MPs, such chatbots may provide a legacy memory – possibly even one manipulated by competitors or by self-learning automated bot-pals or catfishers – of what their constituents’ concerns were during their time as an MP. Time will tell.
CLICK HERE TO DONATE TO OUR CROWDFUNDER
HELP US BECOME STRONGER SO THAT WE CAN CONTINUE TO DELIVER POWERFUL CITIZEN JOURNALISM!
