My Lords, I support this probing amendment, Amendment 251. I thank all noble Lords who have spoken. From this side of the Committee, I say how grateful we are to the noble Lord, Lord Arbuthnot, for all that he has done and continues to do in his campaign to find justice for those sub-postmasters who have been wronged by the system.
This amendment seeks to reinstate the substantive provisions of Section 69 of PACE, the Police and Criminal Evidence Act 1984, revoking this dangerous assumption. I would like to imagine that legislators in 1984 were perhaps alert to the warning in George Orwell’s novel Nineteen Eighty-Four, written some 40 years earlier, about relying on an apparently infallible but ultimately corruptible technological system to define the truth. The Horizon scandal is, of course, the most glaring example of the dangers of assuming that computers are always right. Sadly, as hundreds of sub-postmasters have known for years, and as the wider public have more recently become aware, computer systems can be horribly inaccurate.
However, the Horizon system is very primitive compared to some of the programs which now process billions of pieces of our sensitive data every day. The AI revolution, which has already begun, will exponentially accelerate the risk of compounded errors being multiplied. To take just one example, some noble Lords may be aware of the concept of AI hallucinations. This is a term used to describe when computer models make
inaccurate predictions based on seeing incorrect patterns in data, which may be caused by incomplete, biased or simply poor-quality inputs. In an earlier debate, the noble Viscount, Lord Younger of Leckie, said that account information notices will be decided. How will these decisions be made? Will they be made by individual human beings or by some AI-configured algorithms? Can the Minister share with us how such decisions will be taken?
Humans can look at clouds in the sky or outlines on the hillside and see patterns that look like faces, animals or symbols, but ultimately we know that we are looking at water vapour or rock formations. Computer systems do not necessarily have this innate common sense—this reality check. Increasingly, we will depend on computer systems talking to each other without any human intervention. This will deliver some great efficiencies, but it could lead to greater injustices on a scale which would terrify even the most dystopian science fiction writers. The noble Baroness, Lady Kidron, has already shared with us some of the cases where a computer has made errors and people have been wronged.
Amendment 251 would reintroduce the opportunity for some healthy human scepticism by enabling the investigation of whether there are reasonable grounds for questioning information in documents produced by a computer. The digital world of 2024 depends more on computers than the world of Nineteen Eighty-Four in actual legislation or in an Orwellian fiction. Amendment 251 enables ordinary people to question whether our modern “Big Brother” artificial intelligence is telling the truth when he or it is watching us. I look forward to the Minister’s responses to all the various questions and on the current assumption in law that information provided by the computer is always accurate.