ChatGPT: What are the risks to law firms?

What might it mean for the legal industry if free tools like ChatGPT could be made to generate instantaneous legal documents and draft briefs?

Is allowance instantly strangers applauded

ChatGPT has taken the world by storm. In the few months since it was launched, hundreds of millions of users have experimented with Open AI’s artificial intelligence chatbot to produce copy, answer existential questions, write essays, and generate realistic conversations.

Lawyers and legal firms are anxious, however, about the explosion of interest in ChatGPT. Some firms are asking existential questions of their own, wondering whether AI-driven automation represents risks to their industry – as well as to society at large. 

What might it mean for the legal industry if free tools like ChatGPT could be made to generate instantaneous legal documents and draft briefs? What might come of a computer that can answer complex legal questions, in seconds, with minimal human input? Could developments like ChatGPT be a net positive, and what are the risks to consider?

Data privacy

Legal firms have both a regulatory and a moral obligation to protect their clients’ data. There are numerous policies, regulations, and charters in place to ensure that the highest standards of data privacy and confidentiality are adhered to within the profession.

Partners and their IT teams already go to great lengths to protect this data, using many innovative tools and technologies to ensure that in a digital-first world, sensitive information and communications are protected at all costs. 

But with the rise of AI tools, and ChatGPT in particular, firms will now need to interrogate whether their existing controls are enough. Do they have the necessary processes in place to protect against an AI-related data leak or vulnerability?

ChatGPT users must understand how the tool processes input data, and how inputting any personal or sensitive information could be in violation of data protection laws or industry codes of conduct. Confidentiality or privilege rights are called into question the moment information is inputted into a tool that the user has no control over.

These are the questions that IT teams must be asking themselves when considering how fee earners and partners may use AI language processing tools in the workplace.

Bias and transparency

With any AI tool, the output is dependent on the information with which it is trained, and the people who make decisions about what information the AI receives. This bias is especially apparent in conversations with AI language tools where, for example, a user might ask for an ‘opinion’ on a complex or sticky moral question.

The core issue for firms here is compliance. New regulatory frameworks or codes of conduct may be on the horizon with respect to how AI is used within the legal profession, which would add additional compliance requirements for firms and their lawyers.

Firms also need paper trails and accounts of all work-related conversations they’re having, even if that’s with an AI tool. They also need complete transparency of how and where they gathered the information used to develop any kind of legal document. Without full disclosure and transparency, firms could be opening up themselves to potential compliance or regulatory issues.

While AI tools, including ChatGPT, may provide quick and easy access to a wealth of information – some of which could be useful for lawyers – use of it, for now, still exists in a grey area.

Vulnerabilities

Where there’s a lack of transparency, unfortunately there’s a heightened risk of vulnerability. If a firm or a user doesn’t know exactly who has authority over the tools they’re using – and how these tools function and process data – there’s a risk that vulnerabilities can arise. These vulnerabilities may not be immediately obvious to users or IT teams.

And of course, whenever a firm introduces another tool into its roster of technologies, there’s another potential attack vector to consider and secure.

Already, bad actors have been looking to exploit ChatGPT’s popularity by creating fake apps and tricking users into downloading malware or visiting live phishing page. While these methods of getting users to click on malicious links aren’t new, it’s another consideration for firm partners. If there’s an increase in malicious pages pertaining to an Open AI product, anyone thinking of clicking on those links needs to be especially astute.

Similarly, several cybersecurity firms are warning of ChatGPT’s potential to write compelling phishing emails or malicious code. As previously highlighted, without full transparency into AI language models, attackers could potentially train these tools to generate malicious code.

Ultimately, the responsibility to protect a firm’s data and devices still comes down to the individual firm. But now that there’s another tool to contend with, these responsibilities extend further, and firms must consider adding in additional protections to ensure they don’t get caught out. 

This might take the form of advanced security training for staff or better and more streamlined vulnerability testing. Continuously checking a firm’s security defences against malware and ransomware attacks is increasingly becoming a must-have for all firms within regulated industries, for example.