Top tech lawyer calls for Roman Slave Law and agency to control AI use

One of the UK’s leading legal experts on AI has called for the establishment of seven core principles for the new technology in response the World Ethical Data Foundation’s call for a framework for the safe development of AI products.

Mark Deem, co-author of ‘ AI on Trial’ , a book examining the issues surrounding AI regulation, said it is essential that the implementation of AI is controlled by an agency that governs what it can do, and, where it can be used.

He also pointed out that for AI to be allowed to make financial decisions on our behalf it will be necessary to develop a legal entity to allow it to do so. An entity that could be based upon ancient Roman slave law that empowered slaves who were not Roman citizens to make purchases in Roman shops and markets.

Deem, a partner at the law firm Wiggin, who specialises in digital threats and the opportunities presented by data driven technologies stated that any discussion on regulating the safe deployment of the technology will have to centre on seven key rules that the book sets out.

"During the course of our research for the book we found that there was a need for a set of basic rules that should be laid out to govern AI. The World Ethical Data Forum have laid out the questions that those working on the development of AI models need to ask themselves about developing the technology. What we have done is set out rules that those seeking to control AI can apply to AI systems.

"I think it is remarkable how much the two frameworks complement each other and seek to achieve common goals. These frameworks and ideas do have to be worked on and worked on now, because of the speed at which the technology is being made available. Within the last six months we have seen three of the massive tech giants Microsoft, Google and Meta rolling out AI systems and Elon Musk also announcing plans. We have to bring that activity within a moral and ethical framework."

AI on Trial : the Manifesto

1) Baseline transparency is essential to establishing trust

Those seeking to engage in AI will be required to describe what a given AI product or service is seeking to do; why a decision has been taken to use an automated process; and how such a process occurs.

Through the use of AI Impact Assessments, the potential risks of the technology must be identified at the outset, as well as its impact on the rights and freedoms of individuals, society and the environment.

To achieve this, it will be fundamentally important for those engaging in AI to be able to explain why they have selected particular data sets, how they ensured the accuracy of those data sets and how their algorithms have been coded or devised to use those data sets and how they are protected.

To meet the diversity challenges we have outlined in this book, it is essential that the technology is designed to operate fairly in and across society in general. It must not undermine the protected rights and characteristics of individuals and citizens.

We consider this to be the minimum level of transparency needed to establish trust and unleash the true power of the technology for humanity .

2) An independent supra-national AI oversight agency is vital

We call for significant restrictions to be placed upon the current practice of road-testing so-called 'beta' versions of AI products and services through the creation of an independent supra-national AI oversight agency (Agency).

The role of the Agency will be to assess the potential risks and wider impact of any roll-out of technology identified in AI Impact Assessments, testing where appropriate in a sand-boxed environment. Higher risk applications should necessarily attract higher degrees of scrutiny.

We recommend the introduction of a certification or licensing system to provide assurance to the individual and to establish wider stability of the technology environment. We anticipate that such a system will be adopted by the insurance industry as a pre-requisite of any cover being provided.

To ensure that the wider interests of society and businesses are considered, based on the evidence we have taken, the Agency should as a minimum have representatives safeguarding ethical and philosophical as well as legal considerations.

The Agency should also operate as an advisory body to governments, legislators and regulators alike and as a potential guardian of training qualifications for those developing the technology.

We recommend that it could be funded by contributions from the largest software companies in the world .

3) Ongoing performance monitoring – a black box for the black box

Once certified we recommend that the performance of the technology be monitored through a requirement to maintain logs and an ability of 'interested' third parties or the Agency at suitably determined trigger points to audit and verify performance.

The aim is to create snapshots of the performance of the AI technology in order that, in the event of a system failure and similar to the role of the black box in the aviation industry, investigators can determine what went wrong. In a more than ironic reversal of its linguistic use – the 'black box' in AI refers to where the system carries out calculations at scale and speed and in a manner which might otherwise be entirely impenetrable and incapable of reverse engineering. It is essential that the performance is captured and subject to audit to gain further trust in and ensure that there is greater transparency of the technology in operation.

Further consideration should be given to encouraging and/or requiring companies to provide a statement of compliance of AI technology against its confirmed objectives, in line with the emerging trends of companies on Environmental Social Governance, carbon emissions and diversity statistics .

4) Liability framework to support innovation and improve standards

AI is not sufficiently well-developed or understood to ensure that the declaration of an all embracing AI law, whether nationally or internationally, will succeed or indeed support wider innovation. Rather the focus should be on ensuring that existing laws, policies and procedures are sufficiently flexible to embrace and capture the emergence of new technology.

It is essential that liability is approached in a neutral manner, so that the law responds to an activity carried out by a human in an 'analogue' or 'off-line' environment in the same manner as it would in a 'digital' or 'on-line' environment. Any distinction in approach will serve to undermine the wider liability regime and will lead to evasion or avoidance behaviours.

Whilst we recognise that an AI incorporated into a product might rightly be brought within the strict liability regime of product liability, this is insufficient and inappropriate to cover AI delivered as a service. Accordingly, we consider that fault-based liability be the appropriate standard for assessing liability.

Urgent work should be undertaken to determine whether the development of a liability regime based upon imposition and acceptance of fiduciary duties might align better with the future evolution of the technology .

5) Primacy of interest must be respected, and device sanctity created

In circumstances where a number of different stakeholders have an interest in a device which incorporates AI capability, we need to ensure that any value derived from that device respects and protects the interests of the individual it serves and who requested the use of the AI agent. Devices need to be 'loyal' to the individual who owns them, and the primacy of the individual's interest needs to be protected.

We recommend that there is a need to reinforce what we call 'device sanctity'. Devices considered to have a human interest need to be properly protected against incursions from both state and cyber criminals; a protection enshrined in law.

Consent should be sought from the individual on an opt-in basis if their device (or any of its contents) are to be solicited or used by another party. The implications of such consent should be made clear, including that a technology company might have an active presence on such device. This necessarily goes wider than personal data and the role of Internet of Things devices in people’s houses and their implications will have to adequately spelt out.

6) AI personality and personhood should be limited and exceptional

There are considerable dangers caused by anthropomorphising AI. We should therefore approach any attempt to confer personality onto technology with extreme caution and the lines of demarcation between human personality and legal personality should at all times be clear.

We recognise that there may be limited circumstances where it might be appropriate to consider conferring legal rights on an AI. AI is presently defined and described by computing power and statistics, possibly bounded by some form of encryption. It could, however, be redefined by reference to its creator. These scenarios will be limited and exceptional.

Although AI systems can invent processes and systems, the default position should be that ownership of those processes and systems should be conferred on the AI's owner. The relationship should properly be analogous to the antiquated rights conferred upon slaves to transact on behalf of their masters, which developed towards the end of the Republic in Ancient Rome .

7) Prohibitions

The advent and evolution of new technologies will present use-cases, which should be resisted. We recommend that these should be identified at an early stage with respect of AI. Any encroachment on these areas must be supported by legal sanctions.

At the time of promoting this manifesto, at least three prohibitions should be given strong consideration for moral and ethical reasons.

First, any move to augment human beings physically or fuse them with AI must be strenuously protected. This applies equally to able-bodied as well as disabled individuals.

Secondly, the use of AI in military technology requires careful definition by international standards and must be tightly prescribed if we are going to halt a systematic and gradual erosion of the virtually universal condemnation of the use of AI in warfare.

Thirdly, the deliberate use of algorithmic collusion – where one algorithm effects another algorithm with the intent of either harm or unethical gain to the detriment of individuals and wider society – should be resisted in the absence of some lawful justification.

“These are the principles of a manifesto that we set out for AI in the book a year ago.

“For AI to be safely used and to deliver the benefits that it promises, these seven principles need to be adopted so that we can all be reassured that AI is working in our best interests.”

“It is inevitable that AI will become fully integrated into our lives in the future but to achieve that we have to control that integration so that it occurs in a way that augments us rather than becoming a mechanism of control,” said Deem, adding that transparency was now essential due to the ways that algorithms are already being used in our lives.

By using software that works on predictive methods and seeks to understand our data to accomplish that algorithms were already presenting issues such as the reinforcement of bias said Deem, while efforts to combat that by using algorithms to counter disinformation and fake news could also be interpreted as systematised coercion akin to re-education.

“The world that we are developing is going to be complex. If we are to trust it, then we are going to have to be able to see intent very clearly. That will be on show in ways that we have never seen before, and it will require much more explanation than we are currently seeing with declarations like the cookie policy declarations that we now see.”

A trend to understanding Deem said that will also see the technology community being required to modify its use of language and express itself in plain English to ensure the population at large can understand both technology use and how its data is being used.

“People will have to understand the world that they are getting into they have to be given a significant stake in it,” said Deem.

To arrange an interview with Mark Deem regarding the AI Manifesto please email Blue.buffery@futureintelligence.co.uk

Or call on 07588-878310 and ask for Crawford Warnock.

Top tech lawyer calls for Roman Slave Law and agency to control AI use

Digital twins – creating our own image and making our mark in the metaverse

Cyber crime wave provokes backlash

AI deep fakes fuel fraud

One fifth of the UK wrong side of digital divide and in data poverty claims expert

Insuring the future: the role of cyber insurance in cyber crime

Georgie Warren

Related Posts

Digital twins – creating our own image and making our mark in the metaverse

Cyber crime wave provokes backlash

AI deep fakes fuel fraud

Experts want AGI supranational body

Battle opens with big tech over data use

Top lawyer calls for AI focus

Insuring the future: the role of cyber insurance in cyber crime

Horizon scandal: Claims Law Commission manipulated case for digital evidence

Recent Posts

Categories

Welcome Back!

Create New Account!

Retrieve your password

Are you sure want to unlock this post?

Are you sure want to cancel subscription?

Top tech lawyer calls for Roman Slave Law and agency to control AI use

RELATED POSTS

Support authors and subscribe to content

Subscribe for FREE

One fifth of the UK wrong side of digital divide and in data poverty claims expert

Insuring the future: the role of cyber insurance in cyber crime

Related Posts

Recent Posts

Categories

Welcome Back!

Create New Account!

Retrieve your password

Are you sure want to unlock this post?

Are you sure want to cancel subscription?