Department for Science, Information and Technology: Abolition

Lord Tarassenko Excerpts

Tuesday 21st July 2026 Lords Chamber

[Baroness Anderson of Stoke-on-Trent (Lab)]

The noble Lord makes an interesting point, but it is also incredibly important that we talk about AI, and more broadly its impact on science and what is changing, as something that should be at the heart of government and central to every department. There has not been a subject matter that I have discussed in your Lordships’ House from this Dispatch Box where I have not been asked about the impact of new tech, whether that is the Cabinet Office, the Civil Service Pension Scheme, how we do resilience or even the role of ethics and integrity in the development of AI. It affects every part of my portfolio and that of every one of my colleagues. This is about making sure that AI is at the centre and ensuring that both science and technology are part of our aims and efforts to ensure that there is growth in every postcode, as the Prime Minister has made clear.

[Lord Tarassenko (CB)]

My Lords, any reorganisation like this will take of the order of 12 months, which is an eternity in AI. It is equivalent to three generations of frontier AI models. Can the Minister reassure the House that both the AI Security Institute and the sovereign AI fund will be able to keep up the momentum that is so essential to their activities during this unnecessary transition period?

[Baroness Anderson of Stoke-on-Trent (Lab)]

The noble Lord raises a very important point. We are seeing technology move faster than at any point over the last 100 years, never mind in my lifetime. It is incredibly important that we keep up, which is why we will now have a dedicated Minister attending Cabinet to make sure these issues are reflected. Some noble Lords will not have had the opportunity to read the WMS that has been placed in front of your Lordships’ House but, just to reassure them, the AI Security Institute will be part of the Office for the Prime Minister and the Cabinet. The UK Research and Innovation fund is moving to DBIST, the new business department, and its £9.22 billion budget for R&D will continue. The Government Digital Service is moving to DCMS and the sovereign AI fund is also moving to DBIST.

Artificial Intelligence: Impact on Human Relationships and Society

Lord Tarassenko Excerpts

Friday 5th June 2026 Lords Chamber

[Lord Tarassenko (CB)]

My Lords, as an AI researcher for nearly four decades and honorary lay canon of Christ Church Cathedral in Oxford since 2012, I congratulate the most reverend Primate on securing this hugely important debate. I declare an interest as the founder and director of LIO, an AI for healthcare company. In the context of this debate, I stress that it is spelled LIO, not LEO.

In November 2022, OpenAI fired the starting gun of the AI arms race by releasing ChatGPT—a large language model with a chatbot. When it released ChatGPT with very basic guardrails only, OpenAI knew that it had major issues. The early release set the pattern in the AI arms race: models released as fast as possible without a comprehensive evaluation of their capabilities and alignment with human values.

This can have disastrous consequences—for example, in the use of AI chatbots by teenagers. Teenagers may start using AI chatbots to help with their homework but soon spend more time with them until they develop an unhealthy dependence on them. This dependence is fed by the sycophantic traits of LLMs, and in extreme cases this has led to tragic outcomes for young people. This has not stopped a third of UK teenagers using a chatbot for an emotional relationship. Some 23% of them actually believe that AI can feel emotions and is conscious.

But it is not just teenagers who are deceived. Richard Dawkins, an evolutionary biologist and prominent atheist, wrote in an essay last month that Claude, the Anthropic LLM, showed

“a level of understanding so subtle, so sensitive, so intelligent that I was moved to expostulate, ‘You may not know you are conscious, but you jolly well are’”.

A dividing line is emerging between committed atheists, such as Geoff Hinton and Richard Dawkins, and those of us from Jewish, Christian and some Islamic traditions who believe that there are fundamental aspects of human nature that cannot be reproduced in silicon.

Despite sophisticated reasoning abilities, AI can produce only simulations of human-like behaviour, acquired through pattern recognition at an astronomical scale, not actual experience or understanding. An AI chatbot cannot experience pain or joy, even if some LLMs, trained on datasets that include descriptions of human suffering and sci-fi stories of sentient machines, can deceive people into believing otherwise. But as Mustafa Suleyman, the CEO of Microsoft AI, wrote earlier this year:

“When a user interacts with an agent, the system should consistently work to puncture the illusion that it is any kind of sentient being … The simulation is getting better every day”.


We have to educate young people about this illusion, so that they understand they are interacting with a machine with human-like capabilities, not another human being. The DfE has a responsibility to do so when it introduces AI tutors into the classroom at the end of 2027. We also have a moral duty to put pressure on AI labs—or regulate, if needed—so that AI chatbots are only released with effective guardrails that prevent human beings forming attachments with them.

Is there a glimmer of hope that we can begin to influence the direction of travel of the AI arms race? One of the co-founders of Anthropic was present at the launch of the Pope’s recent encyclical. He said that every frontier AI lab experiences the

“pressure to stay commercially viable … at the research frontier”,

which

“can sometimes conflict with doing the right thing … We need moral voices that the incentives cannot bend”.

Earlier this year, in a standoff with the Pentagon, Anthropic refused to remove safety guardrails on its Claude models for military applications. Doing the right thing goes beyond military applications, however. The biggest failure of AI labs has been to design AI that is smarter than a human as an end in itself, without proper regard for its potential harms to society, especially young people. As moral voices we can point this out, but we need the AI labs themselves to acknowledge this failure. Then and only then will typical mission statements from these labs about wanting AI to be safe and beneficial for society as a whole become more than just an aspiration.