Baroness Fox of Buckley (Non-Afl)

My Lords, I appreciate the reassurances given by the Minister—they are taken into account—but none the less I have two amendments that could go even further. I hope that the Minister can listen. Amendment 50 seeks to ensure that affected parties are informed by their own bank when they are flagged following an eligibility verification notice. To note, the EVM will by design, as has been conceded, detect not only serious fraudsters but people who have been flagged because they have erroneously received an overpayment, possibly due to the DWP’s own mistakes—a possibility not acknowledged enough by measures in this Bill, in my opinion. Because of this, it is important that individuals identified by the power understand the basis on which they have been subjected to greater scrutiny and intrusion. I stress that this amendment is not intended to compromise investigations or tip off fraudsters, as was implied in Committee. In fact, it would come into play only after an individual has been identified by the EVM. This matters because one principle of the rule of law, which I know this Government are very keen on, is that if you are accused of wrongdoing, you should be told the basis for that suspicion. Being flagged by the EVM should not constitute reasonable grounds for suspicion in and of itself, as dealt with in Amendment 52, which I will fully support when we get to it. There is even less reason for an individual who is not suspected of any criminal wrongdoing but who may be the victim of DWP administrative error to be subjected to further checks and requirements without even being told why. It is just not good enough for the DWP to warn people in the personal information charter that the bank account information for the recipients of benefits may be provided to the department. People need to be informed at the moment when the powers bite—that is, once they have been flagged. It is important to note that organisations such as the Public Law Project are keen on seeing this being made a reality. To humanise why this amendment matters, last week several leading charities and advocacy groups, including Big Brother Watch, Age UK and Disability Rights UK, sent a joint letter to DWP Ministers, highlighting the potential human impact and real-life costs of this Bill’s powers. It included testimony from members of the public receiving pension credit, who told their stories to Silver Voices, the advocacy organisation for older people. I was particularly struck by the testimony of Sue from Manchester, who recounted her experience of having to clear her name due to a previous wrongful flag by the DWP, which accused her of having multiple undeclared bank accounts. Sue said, “I had to get letters from each of those banks to clear my name as none of the banks answered the DWP. The mental anguish this has caused me is unbearable and nearly sent me under”. I appreciate that the DWP has gone to great pains to emphasise that it does not consider the EVM to be a government surveillance power, but the idea that you could be flagged by an algorithm that has rifled through your private bank statements and reported you back to the department for further checks, all without you being informed, is—and I say this tentatively—Kafkaesque, so I would like Amendment 50 to correct that. I have also tabled Amendment 62, which requires that the code of practice include scrutiny provisions about the algorithm used by banks—an issue I raised quite a lot in Committee, with some support from fellow Peers. The situation is that the bank accounts of millions of people will be scoured by third-party algorithms that neither the Government nor banks had any hand in creating. It is therefore essential that we be able to scrutinise exactly what these algorithms are doing. I know that the Minister understands the concerns; she has just explained and articulated that well. The Government have said that the code of practice is not the appropriate vehicle for such scrutiny, but we still have not been provided with any alternative. What is more, there has been ministerial opposition to amendments designed to provide greater transparency over the eligibility indicators, which I find worrying. The Minister characterised the algorithm that will be deployed for the EVM as merely comprising two tests that must be met to determine whether an individual can be flagged: whether an account receives a relevant benefit, and whether the account meets the criteria set by the DWP and the EVM. The Minister also said in Committee, and again today, that it is the DWP that will review all the information received and DWP staff who will make any decisions about entitlement where potential fraud or error is identified. But—and it is a big but—that review will only take place on the basis of information returned from the EVM, meaning that we have to be able to scrutinise exactly what is being picked up by the algorithm. They may be simple tests, but they will be applying unpublished eligibility indicators that can complicate issues, depending on what those eligibility indicators require banks to search for. There may be little to scrutinise in algorithms which simply detect whether someone has more than £16,000 in their bank account, or whether, for example, 50% of their transactions have occurred abroad over a period of four weeks. However, should the algorithms have to assess conditions that are more complicated, the algorithms deployed by banks will necessarily be more involved. For example, for relevant benefits currently in scope, such as pension credit, employment and support allowance, and universal credit, could the DWP require banks to find people who they suspect are claiming ESA when not actually suffering from a disability or health condition that affects how much they can work? Or could the DWP require banks to identify recipients of pension credit who live with a partner but claim to live alone? That might be misplaced suspicion, but exactly how the algorithm determines this sort of complex information, and what information it extracts to reach its conclusions, requires oversight. People will otherwise be worried and paranoid; we otherwise risk them being identified for further scrutiny and investigation, potentially on the basis of flawed algorithmic logic. As I emphasised during Committee, determining whether bank accounts meet the eligibility criteria for an EVN requires judgment. It might not be human judgment but algorithmic judgment, yet the question is: what are those criteria? What will the algorithm be asked to search for?