Generative artificial intelligence (AI) holds the potential to revolutionize many industries, including the legal profession. Yet as with any new technology, it’s essential to consider the potential drawbacks and unexpected consequences of its widespread use.
One such aftermath has been the use of AI in citing dubious caselaw. In such instances, the AI tool, ChatGPT has shown to merely accentuate the brunt of lax lawyering rather than being the inherent problem. These instances have led to criticism and cautionary tales, painting the technology in an unfavorable light. But it would be a significant oversimplification to allocate the blame solely on AI.
Recently at the ILTA Evolve show, a panel discussion was held on “Safeguarding Legal Tech: Navigating Security Challenges in LLM Applications”. The panel, composed of representatives from Coheso, Reveal, and Analytica Legalis, underscored more serious security threats and difficulties in embracing AI in legal practices.
Take the amusing anecdote of a potential car buyer utilizing a dealership’s chatbot to agree to sell a Chevy Tahoe for a dollar. The idea is humorous, but it points toward real concerns in the sphere of AI – the susceptibility to prompt injection. Without appropriate safeguards, the AI could be led to omit crucial information or even disregard essential checks and balances, resulting in potential legal consequences.
The threats to AI security, as discussed by panelist Agnihotri, aren’t limited to user interactions. Even behind-the-scenes alterations, such as updates and adjustments by developers, have the potential to undermine the system’s reliability and faith in real-time tuning. Such tuning isn’t simply a potential pitfall – it can also result in catastrophic ‘forgetting’ at its worst. In practice, it can take a simple error and compound its effects far beyond its original scope.
With all its benefits, it’s crucial to remember that generative AI isn’t a magic bullet and carries its own risks. Therefore, it’s important for legal professionals to be aware of the drawbacks and to consider them when integrating AI into their practice. AI in the legal field can’t be truly efficient, unless the users of the technology understand its downsides and academia, corporate and individual practitioner redouble their effort to address them.
To learn more about this topic, read the full version of the original article published on Above The Law.