Should we allow a computer to issue notices on behalf of HMRC's officers? What decisions should computers make? How can we better scrutinise proposed legislation? What safeguards are required?

HMRC has decided that it needs to legislate in order to allow its computers to automatically generate certain types of notice. New proposals for Automated Decisions, follow a string of defeats before the tax tribunal. If enacted this measure will be included in the Finance Act 2020.

Currently, various pieces of legislation set certain safeguards by specifying that ‘an officer, or sometimes a ‘named officer’ must issue certain notices'. If a notice is issued automatically it may thus be invalid. According to the Commissioners for Revenue and Customs Act 2005, an officer is a member of staff and must comply with any directions given by the commissioners of HMRC.

The sort of notices which are intended to be automated, are included in HMRC's 'Automated Decisions: Technical Note' published in October 2019. 

This may well be the 'thin end of the wedge', and it may be good or it may be bad, depending on your attitude to automation.

I am curious to know what new safeguards are to be proposed in transferring responsibilities from staff to computers. Many of the cases which have examined what happens when decisions were not made by officers have arisen as a result of tax appeals under the ‘new’ FA2007-2009 tax penalty regime. That regime of late filing and late payment penalties has proved itself to be anything but the simple ‘tick the box’ system that HMRC promised. Who remembers all those early presentations extolling the easy virtues of the new regime? As the outcomes of so many appeals reveal, we have a right to be sceptical about the merits of any proposed changes that may affect what we file and when we file it.

Without wishing to knock the hard efforts of the various tax bodies who are relied upon by government to scrutinise new legislation, it is clear that ‘desk audits’ have a severe limitation of scope. To start with, many technical writers lack sufficient relevant practical experience to be able to visualise potential outcomes of what happens when unrepresented taxpayers try and work through the system. Then those who work at ‘the coal face’, in public practice, and who could provide the comprehensive walk-through testing that all proposed new legislation so deserves, have day jobs to get on with. There is a limit to what anyone will or is able to do, on a voluntary basis.

For sure, one could write an algorithm which could attempt to sequence all the possible things that a human can think of that would, and could go wrong, but this would still miss something vital. We are only human, after all. Consultation timetables are often short in length and parliamentary scrutiny often nearly non-existent. The result being that only when a piece of legislation is examined by the judiciary that we can discern it true meaning. Often, even then we get it wrong.

Is ‘pass it through parliament first and thoroughly test it later’ a good way to make tax law or any law?

One aspect of the recent High Court ruling on the illegality of police arrests and crowd dispersal of Extinction Rebellion (XR) protesters was that the police commissioner should have served a notice. The XR case is an extreme example, however, we are back to the matter of safeguards. Parliament specifies that humans make certain notices and there is a reason for that.

If the plan is now that computers are making decisions, who programs them, how do we know what has been programmed?

Just a thought.

Nichola Ross Martin CTA (Fellow) FCA, any views expressed here are mine alone.

Hey! Want to share our platform?

If you have a thought or two that you would like express or would like to highlight an issue that really grates. Drop us a line with your contact details and we can add them to the Ten Minute Soapbox.