This article appeared on the online news platform Daily Business and can also be viewed here.
Even since the introduction of ChatGPT, change in the field of AI continues at a breathless pace, bringing with it all manner of challenges in both day-to-day life and in business. That AI will be transformative is indisputable, it already is, but the challenge that remains is how to harness it before it gets out of control.
The recent account of a disgruntled consumer duping a chatbot used by a UK national parcel delivery firm to swear and to compose a poem in which it was highly critical of its “employer”, describing it as useless, is amusing if it is true, or is it just a creation of AI?
Just ask the two lawyers at the New York Bar, who relied upon ChatGPT to assist in their preparations for a court action about the dangers of AI. They had much cause to rue their reliance on ChatGPT when it provided them six decided cases to argue before a judge, none of which existed. They were the product of the programme’s imagination, or hallucination as the unimpressed judge remarked. A large fine for each was the outcome of that misstep, but that must pale into insignificance when compared with the professional embarrassment.
So, it is no wonder that regulators across the world are looking carefully at how to best protect us all from the unintended consequences of AI. Two different approaches are emerging, as some countries favour horizontal AI-specific legislation that applies to all sectors, as in the European Union, and others are choosing to adopt sector-by-sector regulation often based on overall guiding principles which is broadly the approach in the UK.
There has been huge activity in the EU in the last twelve months and that is expected to continue as the EU AI Act comes into force early this year, although many of its provisions will not come into effect immediately. It is an interesting step on the road to regulation and the best example of so-called horizontal legislation. It bans high-risk AIsystems entirely and imposes less stringent requirements for AI systems that are considered less risky.
A key focus is transparency and specifically the requirement to make it clear to a user that they are interacting with an AI programme. The fines for breach have still to be finalised but there has been talk of those being up to €55m or 7% of global turnover. Post Brexit you might think this has nothing to do with the UK but, reflecting the global reach of AI, any UK company which provides AI services in the EU, or uses AI to provide any service there, is caught.
While the EU AI Act focuses on safety, the EU AI Liability Directive establishes a presumption that the AI is the cause of any fault where that seems reasonably likely, and it will also empower courts to order the disclosure of evidence about high-risk AI which is thought to have caused damage.
It is on a similar timetable to the EU AI Act and has the same reach as far as UK companies are concerned. It contrasts with the approach in the UK where the Automated Vehicles Bill, dealing with the apportionment of liability for damages caused by driverless cars, is an example of the sector-by-sector approach.
Time will tell where this ends up but there must be a strong likelihood that there will be a move to standardisation and that the horizontal approach will prevail.
There is much to consider, and away from regulation an important issue in relation to AI is who bears the risk if any third-party copyright is infringed for example, or where data has been scraped from the internet without permission.
In the USA, writer and actor Sarah Silverman has recently raised a law suit against OpenAI and Meta, claiming that her book was used as training material for the AI platforms without permission, without credit, and without payment. And this is an issue for the end user as well as the provider. The lack of transparency and the risk of bias in AI tools used in the workplace is also a cause for concern for employers and workers alike.
There will be a proliferation of laws and regulations, and cases argued in the courts and tribunals on every subject imaginable, over the next decade at least, and businesses would be well advised to take legal advice from that old-fashioned, tried and trusted resource, a human being.