A Guardian reader Q&A on whether artificial intelligence could end humanity surfaced two distinct arguments: whether machines can be conscious, and why the firms building them carry so few binding obligations.
Key takeaways
- The Guardian ran a live question-and-answer session in which its technology reporters responded to readers’ questions about how real the threat from artificial intelligence is, following a week of prominent warnings about the technology.
- One reader asked whether the idea of a humanly conscious AI will ever be dismissed as fantasy, given that human awareness is bound up with senses, emotions, memory and mood.
- The Guardian reports that belief in chatbot consciousness has been voiced by figures far outside AI research, and that in 2022 Google dismissed an employee after claims about a chatbot’s inner life.
- The accountability question raised in the session — why companies issuing existential warnings are not held to account — reflects a genuine gap between the scale of the stated risk and the enforcement mechanisms that currently exist.
The Guardian reader Q&A on AI risk
The item at the centre of this piece is a live blog published by the Guardian in which the paper’s technology reporters took questions from readers about artificial intelligence and existential risk. According to the Guardian, the session followed a week of alarming warnings about AI’s potential to destroy the world as we know it, and was framed around the reality — rather than the rhetoric — of that threat.
The format matters to how the material should be read. A reader Q&A is not an investigation or a peer-reviewed assessment. It is journalists answering questions in near-real time, which produces candid, conversational responses rather than settled conclusions. The questions published, including one from a reader using the handle IcommentthereforeIam, set the agenda as much as the reporters did. Two themes dominate the extract available: whether AI systems can be conscious, and whether the organisations building them face any meaningful consequences.
Claims that chatbots are conscious
The consciousness question put to the Guardian was pointed. Human awareness, the reader argued, is inseparable from sensory input, emotion, memory and shifting mood — none of which a text-generating system has in the ordinary sense. The question was whether the idea of a humanly conscious machine will eventually be accepted as fantasy.
The Guardian’s answer treated this as a serious open question rather than a settled one, noting that the belief has taken hold among people with strong scientific credentials in other fields, including a well-known evolutionary biologist who remains a sceptic on religious questions. The paper also refers to a 2022 case in which Google dismissed an employee in connection with claims about a chatbot. Beyond those two references, no figures, surveys or dates are given in the material available, and it is not known from this source how widespread such beliefs are.
The underlying difficulty is not new, but large language models make it acute. Systems trained on enormous quantities of human text are extremely good at producing language that reads as though it comes from a subject with interests, preferences and continuity of self. That fluency does not, in itself, demonstrate anything about internal experience. Nor does it rule it out: there is no agreed test for consciousness in any system, biological or artificial, and no consensus definition of what would count as evidence. That absence is why the argument persists. A claim that cannot be operationalised cannot be decisively refuted, which is one reason the debate resists the kind of resolution the reader was asking about.
It is worth separating two questions that are often merged. One is whether a model has subjective experience. The other is whether a model behaves in ways that make people treat it as though it does. The second is observable, is already happening, and has consequences for how people use these products, regardless of how the first is eventually answered.
The accountability gap around AI developers
The question quoted in the headline of the Guardian’s session — why these companies are not being held to account — describes a structural feature of the current situation rather than an oversight by any one regulator.
Several things make accountability hard to attach. The harms most often invoked in existential-risk arguments are speculative and future-facing, whereas liability regimes are generally built around demonstrable damage to identifiable parties. The systems themselves are probabilistic and their outputs are not fully predictable from their inputs, which complicates conventional product-safety reasoning. Development is concentrated in a small number of well-resourced firms operating across jurisdictions, so any single national rule reaches only part of the activity. And much of the relevant information — training data, evaluation results, internal safety testing — sits inside the companies, which means external assessment depends substantially on voluntary disclosure.
There is also a rhetorical asymmetry that the reader’s question identifies. A company that publicly warns its own technology could be catastrophically dangerous is, at the same time, continuing to build and sell it. That combination invites the inference that the warnings function partly as marketing for the technology’s power. Whether that inference is correct is not something the Guardian’s material establishes, and it is not knowable from outside. But the tension is real, and it is one reason public statements about existential risk have not translated straightforwardly into public trust.
Existential risk and present-day harm
Running underneath both threads is a disagreement about which risks deserve attention. The existential framing concerns future systems capable enough to act against human interests at scale. A separate body of concern is about systems already deployed: discriminatory outputs, fabricated information presented confidently, labour displacement, surveillance applications, the concentration of computing resources, and the effects of persuasive conversational systems on vulnerable users.
These are not mutually exclusive, but they compete for regulatory attention and for the finite capacity of legislatures. Rules written around hypothetical future capabilities look different from rules written around documented present-day failures, and the two can pull enforcement in different directions. The Guardian’s session, as reported, touched the speculative end. The accountability question a reader asked applies with at least equal force to the documented end, where the evidence of harm is easier to establish and the legal routes are clearer.
What the exchange taken together shows
Read as a whole, the Q&A is less a statement about AI capability than a snapshot of a public conversation that has outrun its evidence base. The consciousness question and the accountability question look unrelated, but they share a root: in both cases, people are being asked to form judgements about systems whose internal workings they cannot inspect, on the basis of behaviour that is designed to be persuasive and claims made by parties with a commercial interest in how the technology is perceived.
That is a difficult position for readers and for regulators alike. It explains why a reader Q&A of this kind attracts the questions it does, and why the honest answers are frequently that something is disputed or not known. The practical implication is that progress on both fronts depends less on resolving the philosophical argument and more on mundane infrastructure: independent evaluation, mandatory disclosure, and legal routes for people harmed by systems already in use. None of that requires agreement on whether a chatbot has an inner life.
Sources and further reading
- The Guardian — the technology desk’s live reader Q&A on AI and existential risk, the primary source for the material described here.
- Academic philosophy of mind literature — for the long-standing problem of testing for consciousness in any system, biological or artificial.
- National and regional AI regulators and standards bodies — for the current state of binding rules, disclosure duties and evaluation requirements.
- AI safety and AI ethics research organisations — for the competing framings of existential risk and documented present-day harm.
Surfaced from the rss:guardian_tech signal “public debate on AI risk”. AI-assisted draft, editorially reviewed.
