'Buy, buy, buy! Adopt, adopt, adopt! Use, use, use!’
The cacophonous sound of AI hype rings in our ears day and night. We’re being stampeded into a world that is not regulated, that is not owned by us (meaning British us), and whose ultimate fate we cannot control. In the main, people who use AI themselves want to use more of it, like an addiction. Those who are on the edge wondering whether to dive in are being urged to take the plunge or miss out.
I won’t deal with whether there is an AI bubble about to burst. It walks and talks like a bubble, but some deny it, and insist it is different to a classic bubble. Who am I to disagree? In any case, I want to talk about the impact on lawyers, and there were two stories this week from the further reaches that might indicate dangers ahead.
First, a journalist heard that one of his stories was going to be adjudicated for truthfulness by an AI tribunal, with the outcome affecting an Honor Index score, a measure of the veracity of published work. A Sackler heir, whose family’s company had been behind the Purdue Pharma opioid scandal, had raised an objection with the AI tribunal over an article that the journalist had written five years previously about him. The headline shouted: ‘A Peter Thiel-Backed Tribunal Is Putting Journalists on Trial. I’m Its First Target’.
The way the tribunal is supposed to work is that a human investigator is assigned to gather evidence. The investigator is cheaper if a college graduate is chosen, more expensive if a former CIA or FBI agent. Then a group of AI models, including well-known ones like Claude, ChatGPT and Grok, act as judge and jury, after analysing the evidence.
The journalist wrote an article about the tribunal. Following his discussions with those behind it, his adjudication was first adjourned, and then the website behind it went dark, saying that due to feedback, it would be rebuilt.
Of course, there has been automated decision-making for some years now (think eBay), but it has been consensual and for small ticket purchases or equivalent. The backers of this AI tribunal compare it to international arbitration centres like the ICC, which is ridiculous.
We are constantly told that we need human-led adjudication and that AI cannot be trusted to give judgements. But I think this story tells us there will be AI tribunals and AI judgements, which will be used as the owners of AI think fit. The further lesson is that, for the time being, there are obviously no proper AI guardrails.
The second story of the week is indeed about guardrails. Since it concerns a decision of the US government, it is not a guardrail that you or I might choose. The US government has not yet legislated for a comprehensive AI regulation package. The picture is patchwork and complex, but in essence there is a policy against states regulating AI on their own, but no successful federal legislation for overall regulation.
There are some things that parts of the US government do not like, and one of them is Anthropic, the makers of Claude, with whom the US government is in dispute. Anthropic has recently released Fable 5 and Mythos 5 which are believed to be able to hunt down and exploit bugs on other systems that have not so far been detected. At the end of last week, Anthropic received a directive from the government to say that it must suspend access by any foreign national to these two new systems. As a result, Anthropic disabled access for all customers, foreign or not. For a full account of the back and forth between the government and Anthropic, see here.
Now, these advanced systems are not designed for law firms, but there is an indirect lesson here about who in the future might be able to use AI, regardless of who is in charge of the US government. There may be no overall regulation, but there are government orders, and foreign nationals might be cut out of usage of a whole range of AI tools. I know I am singing an old song here, but we are entangling ourselves deeply in – and making our justice system dependent on – systems that we don’t own or control.
We have found in national defence that reliance on foreign governments to defend us is unsafe. We are trying to untangle ourselves from that dependence: the difficulties and costs caused the Secretary of State for Defence to resign last week, and might bring down the government.
Why are we doing the same thing with automation, AI and tech, when the consequences are likely to be as grave? (Indeed, defence and AI are becoming inextricably entangled anyway.)
To those shouting ‘Buy! Adopt! Use!’, what is your answer when the US government forbids access by non-US nationals to essential parts of our legal system?
Jonathan Goldsmith is Law Society Council member for EU & International, chair of the Law Society’s Policy & Regulatory Affairs Committee and a member of its board. All views expressed are personal and are not made in his capacity as a Law Society Council member, nor on behalf of the Law Society
No comments yet